Virtual object interaction method, system, device, equipment, medium and program product

Abstract

The embodiment of the invention discloses a virtual object interaction method, system and device, equipment, a medium and a program product. The method comprises the steps that first object information of a target virtual object, second object information of a target role object and outline instruction information are acquired; performing travel outline generation processing on the first object information and the second object information through the target large language model according to an outline generation rule, and generating a travel outline of the target virtual object in a preset time period; acquiring travel description instruction information; performing travel description generation processing on the first object information and the travel outline according to a travel description generation rule through a target large language model, and generating at least one piece of travel description information corresponding to the travel outline; and sending the at least one piece of travel description information to a client corresponding to the target role object. By adopting the embodiment of the invention, the user can be attracted to interact with the virtual object, and the interaction content of the virtual object can be generated in a lightweight manner.

Description

Technical Field

[0001] This application relates to the field of computer application technologies, and in particular, to a method, system, device, equipment, medium, and program product for interacting with virtual objects. Background Art

[0002] With the rapid development of computer and computer network technologies, the Internet has penetrated into all fields of people's work, life, and study. To meet the ever-changing needs of users, virtual objects have emerged. Taking virtual pets as an example of virtual objects, virtual pets have always been popular entertainment and leisure games among users since their birth. Virtual pets are often designed with cute images and, through software support, achieve a growth process similar to that of real pets. With the development of technology, virtual pets already have many advanced functions, and users can obtain rich experiences through pet raising systems, virtual pet communities, and other means.

[0003] Currently, usually, users issue some interaction instructions, and the background generates corresponding interaction content for the virtual pet according to the interaction instructions, thereby attracting users to interact with the virtual object. However, when users do not initiate interaction instructions, the virtual pet will not actively generate corresponding interaction content, so it is difficult to attract users to interact with the virtual pet. In addition, during the "feeding" process of virtual pets, the activity interfaces of virtual pets are mostly presented to users in the form of animations, and users can achieve the "feeding" of virtual pets through the activity interfaces. However, this method of presenting the interaction content of virtual pets through animations requires a large amount of system resources. Therefore, how to attract users to interact with virtual pets and generate the interaction content of virtual pets in a lightweight manner is a technical problem that needs to be solved urgently at present. Summary of the Invention

[0004] The technical problem to be solved by the embodiments of this application is to provide a method, system, device, equipment, medium, and program product for interacting with virtual objects, which can attract users to interact with virtual objects to enhance user viscosity and can generate the interaction content of virtual objects in a lightweight manner.

[0005] In a first aspect, the embodiments of this application provide a method for interacting with virtual objects, and the method includes:

[0006] Obtain the first object information of the target virtual object, the second object information of the target role object, and the outline instruction information; the outline instruction information is used to indicate the outline generation rule;

[0007] Perform outline generation processing on the first object information and the second object information according to the outline generation rule through the target large language model to generate the itinerary outline of the target virtual object within a preset time period;

[0008] Obtain itinerary description instruction information, where the itinerary description instruction information is used to indicate an itinerary description generation rule;

[0009] Perform itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rule through the target large language model, and generate at least one itinerary description information corresponding to the itinerary outline;

[0010] Send the at least one itinerary description information to the client corresponding to the target role object, and the client is used to display the at least one itinerary description information.

[0011] In a second aspect, an embodiment of the present application provides an interaction system for virtual objects. The system includes a client and a server. The server includes an itinerary service module and an itinerary generation module, where:

[0012] The client is used to send an itinerary acquisition request to the server;

[0013] The itinerary service module is used to, in response to the itinerary acquisition request, obtain at least one itinerary description information from the itinerary generation module, and send the at least one itinerary description information to the client;

[0014] The client is further used to display at least one itinerary description information;

[0015] The itinerary generation module is used to perform itinerary outline generation processing on the first object information and the second object information according to the outline generation rule through the target large language model, generate an itinerary outline of the target virtual object within a preset time period, and perform itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rule through the target large language model, and generate at least one itinerary description information corresponding to the itinerary outline.

[0016] In a third aspect, an embodiment of the present application provides an interaction device for virtual objects. The device includes:

[0017] An acquisition unit, configured to acquire first object information of a target virtual object, second object information of a target role object, and outline instruction information; the outline instruction information is used to indicate an outline generation rule;

[0018] A generation unit, configured to perform itinerary outline generation processing on the first object information and the second object information according to the outline generation rule through the target large language model, and generate an itinerary outline of the target virtual object within a preset time period;

[0019] The acquisition unit is further configured to acquire itinerary description instruction information, where the itinerary description instruction information is used to indicate an itinerary description generation rule;

[0020] The generating unit is further configured to perform a travel description generation process on the first object information and the travel outline according to the travel description generation rule through the target large language model, so as to generate at least one travel description information corresponding to the travel outline.

[0021] The sending unit is configured to send the at least one travel description information to the client corresponding to the target role object, and the client is configured to display the at least one travel description information.

[0022] In a fourth aspect, an embodiment of the present application provides a computer device, which includes a memory, a communication interface, and a processor. Among them, the memory, the communication interface, and the processor are interconnected; the memory stores a computer program, and the processor calls the computer program stored in the memory to implement the interactive method of the virtual object in the first aspect above.

[0023] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program, and the computer program, when executed by a processor, implements the interactive method of the virtual object in the first aspect above.

[0024] In a sixth aspect, an embodiment of the present application provides a computer program product, which includes a computer program stored in a computer storage medium; a processor of a computer device reads the computer program from the computer storage medium, and the processor executes the computer program to enable the computer device to execute the interactive method of the virtual object in the first aspect above.

[0025] In the embodiments of the present application, by obtaining the first object information of the target virtual object, the second object information of the target role object, and the outline instruction information; the outline instruction information is used to indicate the outline generation rule; by using the target large language model to perform outline generation processing on the first object information and the second object information according to the outline generation rule, an itinerary outline of the target virtual object within a preset time period is generated; obtaining itinerary description instruction information, which is used to indicate the itinerary description generation rule; by using the target large language model to perform itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rule, at least one itinerary description information corresponding to the itinerary outline is generated; sending the at least one itinerary description information to the client corresponding to the target role object, and the client is used to display the at least one itinerary description information. The embodiments of the present application generate the itinerary outline and itinerary description information in a hierarchical manner, gradually refine the itinerary content, and combine the generation capabilities of the large language model for different tasks, and can generate the itinerary description information of the target virtual object with text content at a relatively low cost, realizing the lightweight generation of interactive content of the virtual object. In addition, even when the user does not initiate an interaction instruction, the virtual pet will actively generate corresponding interactive content, which can attract the user to interact with the virtual object to enhance user stickiness. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the background art, the following will describe the drawings required to be used in the embodiments of the present application or the background art.

[0027] Figure 1 It is a schematic structural diagram of an interactive system for virtual objects provided by an embodiment of the present application;

[0028] Figure 2 It is a schematic flowchart of an interactive method for virtual objects provided by an embodiment of the present application;

[0029] Figure 3A It is a schematic diagram of an itinerary description interface provided by an embodiment of the present application;

[0030] Figure 3B It is a schematic diagram of an itinerary description interface provided by an embodiment of the present application;

[0031] Figure 3C It is a schematic diagram of an itinerary track interface provided by an embodiment of the present application;

[0032] Figure 3D It is a schematic diagram of an itinerary track interface provided by an embodiment of the present application;

[0033] Figure 3E It is a schematic diagram of an itinerary track interface provided by an embodiment of the present application;

[0034] Figure 3FIt is a schematic diagram of the itinerary description interface provided by an embodiment of the present application;

[0035] Figure 3G It is a schematic diagram of the itinerary description interface provided by an embodiment of the present application;

[0036] Figure 4 It is a schematic flowchart of another method for interacting with virtual objects provided by an embodiment of the present application;

[0037] Figure 5 It is a management architecture diagram of virtual objects provided by an embodiment of the present application;

[0038] Figure 6 It is a generation architecture diagram of a kind of itinerary description information provided by an embodiment of the present application;

[0039] Figure 7 It is a schematic structural diagram of an interaction device for virtual objects provided by an embodiment of the present application;

[0040] Figure 8 It is a schematic structural diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0041] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0042] At the same time, in the description of the embodiments of the present application, terms such as "first" and "second" are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more features. In the description of the embodiments of the present application, the meaning of "a plurality" is two or more, unless otherwise specifically defined.

[0043] In the specific implementation manner of the present application, data related to users, such as basic information of users, session messages, etc., need to obtain user permission or consent when the embodiments of the present application are applied to specific products or technologies, and the collection, use, and processing of relevant data need to comply with local laws, regulations, and standards.

[0044] The embodiments of the present application provide an interaction system for virtual objects. Exemplarily, as Figure 1 shown, Figure 1It is a schematic structural diagram of an interaction system for virtual objects provided by an embodiment of the present application. The interaction system for virtual objects may include at least one client 100 and a server 200. A computer-readable storage medium corresponding to the interaction method for virtual objects is run in the server 200 to execute the interaction method for virtual objects. The server 200 communicates with at least one client 100 and is used to send itinerary description information to any client 100.

[0045] Among them, the server 200 may include an itinerary service (such as Figure 1 the world display service shown) module and an itinerary generation (such as Figure 1 the world log simulation service shown) module. The client 100 is used to send an itinerary acquisition request to the server 200; the itinerary service module is used to respond to the itinerary acquisition request, obtain at least one itinerary description information from the itinerary generation module, and send the at least one itinerary description information to the client 100; the client 100 is also used to display the at least one itinerary description information; the itinerary generation module is used to perform itinerary outline generation processing on the first object information and the second object information according to the outline generation rules through the target large language model, generate an itinerary outline of the target virtual object within a preset time period, and perform itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rules through the target large language model to generate at least one itinerary description information corresponding to the itinerary outline. Optionally, the server 200 may further include an itinerary database (such as Figure 1 the world log database shown), and the itinerary database is used to store the itinerary outline, itinerary description information, and itinerary details information generated by the itinerary generation (world log simulation service) module. Optionally, the server 200 may further include a memory database, and the memory database is used to store historical session messages and historical memory information.

[0046] Specifically, Object 1 can view the itinerary description information of the target virtual object for one day through the interaction interface of Client 100, and can trigger each itinerary description information to view more detailed itinerary details, such as the mood, thoughts, behaviors, and surrounding environment of the target virtual object. The itinerary description information and itinerary details information can be generated by the World Log Simulation Service. Among them, an itinerary outline is generated by the outline generator in the World Log Simulation Service. The outline generator can send an itinerary outline generation request to the Large Language Model (LLM) service. The target LLM obtains the first object information of the target virtual object, the second object information of the target role object (Object 1), and the outline instruction information. The outline instruction information is used to indicate the outline generation rule. The target LLM performs itinerary outline generation processing on the first object information and the second object information according to the outline generation rule, generates the itinerary outline of the target virtual object within a preset time period, and the LLM service replies to the request and sends the itinerary outline to the outline generator. Among them, the itinerary description information is generated by the planning generator in the World Log Simulation Service. The planning generator can send an itinerary description information generation request to the LLM service. The target LLM obtains the itinerary description instruction information, and the itinerary description instruction information is used to indicate the itinerary description generation rule. The target LLM performs itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rule, generates at least one itinerary description information corresponding to the itinerary outline, and the LLM service replies to the request and sends the at least one itinerary description information to the planning generator. Among them, the itinerary details information is generated by the detail generator in the World Log Simulation Service. The detail generator can send an itinerary details information generation request to the LLM service. The target LLM obtains the itinerary details instruction information, and the itinerary details instruction information is used to indicate the itinerary details generation rule. The target LLM performs itinerary details generation processing on the first object information, historical memory information, itinerary outline, and any itinerary description information according to the itinerary details generation rule, generates the itinerary details information corresponding to any itinerary description information, and the LLM service replies to the request and sends the itinerary details information to the detail generator. Optionally, if the scene of the target virtual object is a memory scene, the World Log Simulation Service will also send a memory request to the memory database, so that the memory database searches for historical memory information and sends the historical memory information (user memory) to the World Log Simulation Service to generate the itinerary outline, itinerary description information, and itinerary details information in combination with the historical memory information.

[0047] Among them, the server 200 can store the itinerary outline, itinerary description information, and itinerary details information generated from the world log in the world log database in a daily update manner. When the object 1 interacts with the target virtual object through the client 100, the client 100 sends a user request to the world display service. The world display service requests the itinerary description information for the corresponding date from the world log database according to the date of the itinerary of the object 1 indicated by the user request. The world log database sends the itinerary description information (such as Figure 1 the daily world information shown) to the world display service. The world display service performs a certain degree of display processing on the itinerary description information (daily world information) and sends the itinerary description information to the client 100 for display. Optionally, when the object 1 triggers any itinerary description information, the world display service will also obtain the corresponding itinerary details information through the world log database and perform display processing.

[0048] Optionally, the server 200 may further include a dialogue service module and / or a memory summary module, where: the dialogue service module is used to, in response to a dialogue reply request, obtain historical memory information and an itinerary outline, and generate a reply message for the target virtual object by performing reply message generation processing on the first object information, second object information, context information, historical memory information, and itinerary outline through a target large language model according to the reply message generation rule; the memory summary module is used to perform memory extraction processing on the conversation messages within a preset historical time period through the target large language model according to the memory extraction rule to generate historical memory information.

[0049] Specifically, after the object 1 views the itinerary description information or itinerary details information of the target virtual object, if an interaction idea with the target virtual character is generated, a dialogue can be conducted with the target virtual object through a dialog box. During the dialogue process, the client 100 sends a user request to the dialogue service in the server 200. The dialogue service sends a request to the world log database to obtain the daily world information and sends a memory request to the memory database to obtain the memory of the user (object 1). The dialogue service requests to call the LLM service to process the daily world information (itinerary description information), user memory (historical memory information), first object information, second object information, and context information to generate a reply message. The LLM service replies to the dialogue service and sends the reply message to the dialogue service. The dialogue service will process the reply message into a suitable form and send the dialogue reply (reply message) to the client 100 for display.

[0050] Specifically, after the target character object (Object 1) has a conversation with the target virtual object, the conversation content (historical conversation messages) will be temporarily stored in the memory database. The memory database will send a memory request to the memory summarization service. The memory summarization service will read it for each conversation and request to call the LLM service to perform memory extraction processing on the conversation messages within a preset historical time period according to the memory extraction rules, generating historical memory information. The LLM service will reply to the memory summarization service and send the historical memory information to the memory summarization service. The memory summarization service will clean the historical memory information in a reasonable format and finally send it to update the memory in the memory database.

[0051] Among them, in the overall implementation framework, the world log simulation service and the memory summarization service are both called daily. For a single user, the memory summarization service will be called according to the number of conversation rounds, while the world log simulation service will call the LLM according to one itinerary outline, one itinerary description information call, and no more than 10 calls (entries) for the itinerary refinement process. In this way, the service pressure can be transferred to the low-frequency interval of the conversation service, reducing the service call pressure during peak periods. In addition, the conversation service, the memory database, the LLM service, the world log database, and the world display service are all online; the memory summarization service and the world log simulation service are offline.

[0052] The server 200 in the interaction system of the virtual object can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0053] Any client 100 and the server 200 can be directly or indirectly communicatively connected through wired or wireless communication methods, and this application does not limit this here.

[0054] It can be understood that Figure 1 The server in the interaction system of the virtual object shown, or at least one client, does not constitute a limitation to the embodiments of this application. That is, the number of servers, the types of servers, the number of at least one client, the types of at least one client, or the number of devices and the types of devices included in each server and client do not affect the overall implementation of the technical solutions in the embodiments of this application, and can all be regarded as equivalent replacements or derivatives of the technical solutions required to be protected in the embodiments of the present invention.

[0055] Based on the above description, please refer to Figure 2 ,Figure 2 It is a schematic flowchart of an interaction method for virtual objects provided by an embodiment of the present application. This interaction method for virtual objects can be applied in a server. For example, Figure 2 the interaction method for virtual objects shown includes but is not limited to steps S201 - S205, where:

[0056] S201, obtain the first object information of the target virtual object, the second object information of the target role object, and the outline instruction information; the outline instruction information is used to indicate the outline generation rule.

[0057] In this embodiment, the virtual object can include a virtual pet, a virtual creature, or a virtual element, etc. A virtual pet can refer to a pet in the virtual world, such as a dog or a cat existing in a virtual object application. A virtual creature can refer to an object with vital signs in the virtual world, such as an animal or a plant existing in a virtual object application. A virtual element can refer to an element in the virtual world, such as an object with vital signs or an object without vital signs. An object without vital signs can be, for example, a house, a tire, etc. The role object can be a player (Player), such as a game participant. The first object information can include the preference information, emotion information, identity information, and attribute information of the virtual object. The preference information of the virtual object can include the interests and hobbies, personal habits, etc. of the virtual object. The preference information of the virtual object can be gradually optimized during the interaction with the user, so as to obtain a virtual object whose preference information meets the user's expectations, getting closer to the identity of the user's "friend", thereby establishing a deeper emotional bond between the user and the pet. The identity information of the virtual object can be preset, such as the type, weight, gender, age, etc. The attribute information of the virtual object is used to indicate the object behavior of the virtual object, such as the action of "wagging the tail". The second object information can include the preference information, identity information, etc. of the role object (user). The outline generation rule can include the word count rule of the outline and the content rule of the outline, etc. For example, if the outline instruction information is to generate a travel outline of the target virtual object based on the first object information and the second object information, with no more than 30 words, then the word count rule of the outline is no more than 30 words, and the content rule of the outline is a travel outline related to the first object information and the second object information.

[0058] Optionally, the outline instruction information can be input by the administrator corresponding to the server at a preset time. For example, the administrator inputs the outline instruction information at zero o'clock every day, so that the target large language model performs travel outline generation processing on the first object information and the second object information according to the outline generation rule indicated by the outline instruction information, and generates a travel outline of the target virtual object within a preset time period; the outline instruction information can also be automatically generated through a trigger mechanism. The trigger mechanism can be a preset trigger time. For example, when it reaches zero o'clock every day, the server will automatically generate the outline instruction information.

[0059] S202. Generate a travel itinerary for the target virtual object within a preset time period by performing travel itinerary generation processing on the first object information and the second object information according to the itinerary generation rules of the target large language model.

[0060] In this embodiment, the target large language model performs travel itinerary generation processing on the first object information and the second object information according to the itinerary generation rules indicated by the itinerary instruction information, such as the word count rule of the itinerary and the content rule of the itinerary, to generate a travel itinerary for the target virtual object within a preset time period. For example, the first object information and the second object information can be subjected to feature vector extraction through the embedding layer of the target large language model. The encoder layer understands the input feature vectors through the self-attention mechanism. The decoder layer interprets the required information based on the input feature vectors and combines the self-attention and multi-head attention mechanisms, and finally gradually generates each element of the travel itinerary (such as the goals and activities for each day).

[0061] Among them, the generation of the travel itinerary combines the first object information and the second object information, so the content of the travel itinerary will include the interaction between the target virtual object and the target role object.

[0062] In an alternative embodiment, it is also possible to obtain the fourth object information of the training virtual object, the fifth object information of the training role object, the preset travel itinerary, and the itinerary instruction information; the itinerary instruction information is used to indicate the itinerary generation rules; perform travel itinerary generation processing on the fourth object information and the fifth object information according to the itinerary generation rules through the initial large language model to generate a travel itinerary for the training virtual object within a preset time period; train the initial large language model with the goal of reducing the difference between the travel itinerary of the training virtual object within the preset time period and the preset travel itinerary, and obtain the target large language model after the training is completed; the target large language model is used to perform travel itinerary generation processing on the first object information and the second object information according to the itinerary generation rules to generate a travel itinerary for the target virtual object within a preset time period. By training the target large language model through the above method, the target large language model can accurately generate a travel itinerary according to the first object information and the second object information.

[0063] In an alternative embodiment, world rule information can also be obtained; the target large language model performs itinerary outline generation processing on the first object information, the second object information, and the world rule information according to the outline generation rules indicated by the outline instruction information, and generates an itinerary outline of the target virtual object within a preset time period. Among them, the world rule information is used to indicate the world rules of the virtual environment where the target virtual object is located. The world rule information is the rules defined in the virtual object application. The world rules of the virtual environment where the target virtual object is located can be used to indicate the daily behaviors and itinerary content of the target virtual object, etc. For example, if the target virtual object is a virtual pet, then the virtual pet does not need to work every day; another example is that the virtual pet is equal to the player user, and the virtual pet cannot verbally attack the player user, etc.

[0064] In an alternative embodiment, the target large language model performs itinerary outline generation processing on the first object information and the second object information according to the outline generation rules to generate an itinerary outline of the target virtual object within a preset time period. This can be achieved through the following method: the target large language model performs itinerary outline processing on at least one of the first object information, the second object information, the historical memory information, and the social relationship between the target virtual object and other virtual objects according to the outline generation rules indicated by the outline instruction information, and generates an itinerary outline of the target virtual object within a preset time period; the historical memory information is obtained based on the historical conversation messages between the target virtual object and the target role object.

[0065] In this embodiment, other virtual objects can refer to virtual objects different from the target virtual object in the virtual environment where the target virtual object is located, that is, non-player characters (NPCs). Optionally, a setting library containing a preset number of NPCs can be created. For example, a setting library containing 100 NPCs is created. By combining the third object information of any NPC with the first object information of the target virtual object, the social relationship between the target virtual object and other virtual objects is obtained. The target large language model performs itinerary outline generation processing on the first object information, the second object information, the historical memory information, and the social relationship according to the outline generation rules, and generates an itinerary outline of the target virtual object within a preset time period. Among them, the method of combining the third object information of any NPC with the first object information of the target virtual object can be random combination. The outline content of the itinerary outline will include the interaction between the target virtual object and the NPC. By randomly combining object information, a social network is customized for the target virtual object, so as to simulate a complete community structure at a lower cost and add the social attributes and itinerary life richness of the target virtual object.

[0066] In an alternative embodiment, multiple scenario types can also be preset for the target virtual object. The preset multiple scenarios can include a memory scenario and a non-memory scenario. The non-memory scenario can include a normal scenario and an abnormal scenario. Among them, the normal scenario can refer to that when generating the itinerary outline of the target virtual object, historical memory information is not required, and when generating the itinerary outline through the large language model, only partial second object information is input, so that the generated itinerary outline is more independent, that is, the personalized itinerary outline of the target virtual object; the abnormal scenario can refer to that when generating the itinerary outline of the target virtual object, historical memory information is not required, and the target role object and the target virtual object do not interact within the preset interaction time. At this time, the goal of generating the itinerary outline is to enable the target role object and the target virtual object to interact; the memory scenario can refer to that when generating the itinerary outline of the target virtual object, historical memory information is required.

[0067] In an alternative embodiment, the target scenario type of the target virtual object can also be selected from the preset multiple scenario types; if the target scenario type is a memory scenario, historical memory information is obtained; the historical memory information is obtained based on the historical conversation messages between the target virtual object and the target role object; the first object information, the second object information, and the historical memory information are processed for the itinerary outline by the target large language model according to the outline generation rules to generate the itinerary outline of the target virtual object within the preset time period. Among them, the target scenario type of the target virtual object can be selected from the preset multiple scenario types by means of random selection or specified selection. The outline instruction information corresponding to different target scenario types is different. The outline instruction information of the memory scenario can be I 1m , then the itinerary outline of the memory scenario can be expressed by formula (1) as:

[0068] T 1m = φ(I 1m , P p , P u , N, M) (1)

[0069] Among them, T 1m represents the itinerary outline of the memory scenario, I 1m represents the outline instruction information of the memory scenario, P p represents the first object information, P u represents the second object information, N represents the third object information, and M represents the historical memory information.

[0070] If the target scenario type is a non-memory scenario, the target large language model is used to perform itinerary outline generation processing on the first object information and the second object information according to the outline generation rules, and an itinerary outline of the target virtual object within a preset time period is generated. The non-memory scenario can include a normal scenario and an abnormal scenario. When the target role object and the target virtual object do not interact within the preset interaction time, it can be determined that the target scenario type of the target virtual object is an abnormal scenario. The outline instruction information of the normal scenario can be I 1n , then the itinerary outline of the normal scenario can be expressed by formula (2) as:

[0071] T 1n = φ(I 1n , P p , P u , N) (2)

[0072] Among them, T 1n represents the itinerary outline of the normal scenario, I 1n represents the outline instruction information of the normal scenario, P p represents the first object information, P u represents the second object information, and N represents the third object information.

[0073] The outline instruction information of the abnormal scenario can be I 1a , then the itinerary outline of the abnormal scenario can be expressed by formula (3) as:

[0074] T 1a = φ(I 1a , P p , P u , N) (3)

[0075] Among them, T 1a represents the itinerary outline of the abnormal scenario, I 1a represents the outline instruction information of the abnormal scenario, P p represents the first object information, P u represents the second object information, and N represents the third object information.

[0076] Among them, I 1n , I 1a , I 1m , P p , P u , M, N, T 1n , T 1a , T 1m are all text sequences in the form of {t1, t2, …, t n}, where t i , i ∈ {1, 2, …, n} is a single character, and n is the length of each string.

[0077] By designing different scenarios to generate itinerary outlines corresponding to the scenarios, the generated itinerary outlines can better conform to the user's interests and wishes, attract the user to interact with the target virtual object, ensure the user experience, and increase user viscosity.

[0078] S203. Obtain itinerary description instruction information, which is used to indicate the itinerary description generation rule.

[0079] In this embodiment, the itinerary description generation rule may include the word count rule of the itinerary description, the number rule of the itinerary description, and the content rule of the itinerary description, etc. For example, the itinerary description instruction information is to generate 5 to 10 itinerary description information corresponding to the itinerary outline according to the first object information, the second object information, and the itinerary outline, and each itinerary description information does not exceed 10 words. Then the word count rule of the itinerary description is not to exceed 10 words, the number rule of the itinerary description is not less than 5 and not more than 10 itinerary description information, and the content rule of the itinerary description is the itinerary description information related to the first object information, the second object information, and the itinerary outline.

[0080] Optionally, the itinerary description instruction information may be input by the administrator corresponding to the server at a preset time. For example, the administrator inputs the itinerary description instruction information at zero o'clock every day, so that the target large language model performs itinerary description generation processing on the first object information, the second object information, and the itinerary outline according to the itinerary description generation rule, and generates itinerary description information corresponding to the itinerary outline; the itinerary description instruction information may also be automatically generated through a trigger mechanism, and the trigger mechanism may be a preset trigger time. For example, when it reaches zero o'clock every day, the server will automatically generate the itinerary description instruction information.

[0081] Optionally, the outline instruction information and the itinerary description instruction information may be input together by the administrator corresponding to the server. For example, the outline instruction information and the itinerary description instruction information are input together. When the server needs to generate an itinerary outline, it obtains the outline instruction information. When the server needs to generate itinerary description information, it obtains the itinerary description instruction information. Optionally, the outline instruction information and the itinerary description instruction information may be input separately by the administrator corresponding to the server. For example, when the server is required to generate an itinerary outline, the outline instruction information is input. When the server is required to generate itinerary description information, the itinerary description instruction information is input.

[0082] S204. Through the target large language model, perform itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rule, and generate at least one itinerary description information corresponding to the itinerary outline.

[0083] In this embodiment, the target large language model performs itinerary description generation processing on the first object information and the itinerary outline according to itinerary description generation rules, such as the word count rule of the itinerary description, the number rule of the itinerary description, and the content rule of the itinerary description, to generate at least one itinerary description information corresponding to the itinerary outline. For example, the first object information and the itinerary outline can be subjected to feature vector extraction through the embedding layer of the target large language model. The encoder layer uses the self-attention mechanism to understand the input feature vectors. The decoder layer interprets the required information based on the input feature vectors and combines self-attention and multi-head attention mechanisms, and finally gradually generates each element of each itinerary description information (such as the itinerary description content for each hour). Optionally, the target large language model can also perform itinerary description generation processing on the first object information, the second object information, and the itinerary outline according to the itinerary description generation rules to generate at least one itinerary description information corresponding to the itinerary outline. Among them, the first object information and the second object are input into the target large language model again to ensure that the itinerary description information generated by the large language model does not violate common sense.

[0084] In an alternative embodiment, it is also possible to obtain the fourth object information of the training virtual object, the preset itinerary outline, the preset itinerary description information, and the itinerary description instruction information; the itinerary description instruction information is used to indicate the itinerary description generation rules; the initial large language model performs itinerary description generation processing on the fourth object information and the preset itinerary outline according to the itinerary description generation rules to generate at least one itinerary description information corresponding to the preset itinerary outline; the initial large language model is trained with the goal of reducing the difference between at least one itinerary description information corresponding to the preset itinerary outline and the preset itinerary description information, and the target large language model is obtained after the training is completed; the target large language model is used to perform itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rules to generate at least one itinerary description information corresponding to the itinerary outline. By training the target large language model through the above method, the target large language model can accurately generate itinerary description information according to the first object information and the itinerary outline. Optionally, it is also possible to obtain the fourth object information of the training virtual object, the fifth object information of the training role object, the training itinerary outline, the preset itinerary description information, and the itinerary description instruction information to train the initial large language model. For the specific training process, please refer to the above description, and this solution will not be elaborated here.

[0085] In an alternative embodiment, it is also possible to obtain world rule information; the target large language model performs itinerary description generation processing on the first object information, the second object information, the itinerary outline, and the world rule information according to the itinerary description generation rules to generate at least one itinerary description information corresponding to the itinerary outline.

[0086] In an alternative implementation, it is possible to obtain the consensus information and content order of the itinerary content corresponding to each itinerary description information in at least one itinerary description information. Each itinerary description information in the at least one itinerary description information includes at least the itinerary time and itinerary content. For example, one of the itinerary description information is "Xiaopa goes to school at 10 o'clock". The itinerary time included in each itinerary description information is determined according to the consensus information of the itinerary content corresponding to each itinerary description information and the content order of the itinerary content included in the at least one itinerary description information. The consensus information refers to the information of the common cognition and acceptance state of a certain problem, rule or value formed by social members through interaction and coordination. For example, one of the itineraries is "Xiaopa has breakfast", and the other is "Xiaopa goes to school". Then the action of having breakfast should be before going to school. Then it can be determined that one of the itinerary description information is "Xiaopa has breakfast at 9 o'clock". According to the consensus information, the time for having breakfast is about 20 minutes. Then it can be determined that the other itinerary description information is "Xiaopa goes to school at 9:20". Then, the target large language model can perform itinerary description generation processing on the first object information, consensus information, content order, and itinerary outline according to the itinerary description generation rules to generate at least one itinerary description information corresponding to the itinerary outline.

[0087] In an alternative implementation, it is also possible to obtain the attribute information of the target virtual object. The attribute information is used to indicate the object behavior of the target virtual object. The first object information includes the attribute information. Then, the target large language model can perform itinerary description generation processing on the first object information, attribute information, and itinerary outline according to the itinerary description generation rules to generate at least one itinerary description information corresponding to the itinerary outline. Among them, inputting the attribute information of the target virtual object into the target large language model is to ensure that the itinerary description information generated by the large language model does not deviate from the behavior of the target virtual object. For example, if the target virtual object is a virtual pet dog, the attribute information of the virtual pet dog will indicate the virtual pet dog to perform the action of "wagging its tail". For example, if the target virtual object is a virtual pet bird, the attribute information of the virtual pet bird will indicate the virtual pet bird to perform the action of "preening its feathers".

[0088] Optionally, it is also possible to perform itinerary description generation processing on the first object information, consensus information, content order, consensus information, and itinerary outline by the target large language model according to the itinerary description generation rules to generate at least one itinerary description information corresponding to the itinerary outline. For the specific steps of this solution, please refer to the above description, and this solution will not be elaborated here.

[0089] Optionally, if the target scenario type is a memory scenario, then the target large language model can perform itinerary description generation processing on the first object information, the second object information, the third object information, the historical memory information, and the itinerary outline according to the itinerary description generation rules, and generate at least one itinerary description information corresponding to the itinerary outline. The itinerary description instruction information corresponding to different target scenario types is different. The itinerary description instruction information for the memory scenario can be I 2m , then the itinerary description information for the memory scenario can be expressed by formula (4) as:

[0090] {T 2mi} = φ(I 2m , T 1m , P p , P u , N, M), i ∈ {1, 2, …, K3} (4)

[0091] Among them, {T 2mi} represents the K3 generated itinerary description information, I 2m represents the itinerary description instruction information for the memory scenario, T 1m represents the itinerary outline for the memory scenario, P p represents the first object information, P u represents the second object information, N represents the third object information, and M represents the historical memory information.

[0092] Optionally, if the target scenario type is a non-memory scenario, then the target large language model performs itinerary description generation processing on the first object information, the second object information, the third object information, and the itinerary outline according to the itinerary description generation rules, and generates at least one itinerary description information corresponding to the itinerary outline. The non-memory scenario can include a normal scenario and an abnormal scenario. When there is no interaction between the target role object and the target virtual object within the preset interaction time, it can be determined that the target scenario type of the target virtual object is an abnormal scenario. The itinerary description instruction information for the normal scenario can be I 2n , then the itinerary description information for the normal scenario can be expressed by formula (5) as:

[0093] {T 2ni} = φ(I 2n , T 1n , P p , P i , N), i ∈ {1, 2, …, K1} (5)

[0094] Among them, {T 2ni} represents the K1 generated itinerary description information, I 2n represents the itinerary description instruction information for the normal scenario, T 1n represents the itinerary outline for the normal scenario, P pRepresents the first object information, P u Represents the second object information, and N represents the third object information.

[0095] The itinerary description instruction information for the abnormal scenario can be I 2a Then, the itinerary description information for the abnormal scenario can be expressed by formula (6) as:

[0096] {T 2ai} = φ(I 2a , T 1a , P p , P u , N), i ∈ {1, 2, …, K2} (6)

[0097] Wherein, {T 2ai} represents the generated K2 itinerary description information, I 2a represents the itinerary description instruction information for the normal scenario, T 1a represents the itinerary outline for the abnormal scenario, P p represents the first object information, P u represents the second object information, and N represents the third object information.

[0098] By designing different scenarios to generate the itinerary description information corresponding to the scenarios, the generated itinerary description information can be made more in line with the user's interest and willingness, attracting the user to interact with the target virtual object, ensuring the user experience, and enhancing user viscosity.

[0099] In an alternative implementation, it is also possible to determine the expression mark corresponding to any itinerary description information in at least one itinerary description information according to the mapping relationship between the itinerary description information and the expression mark; send the expression mark to the client, and the client is used to display the expression mark. Among them, the itinerary description information and the expression mark can be matched through a matching model to determine the expression mark corresponding to each itinerary description information. Displaying the expression mark on the client can make the interface of the client simple. The expression mark can be an emoji. Optionally, the expression mark description information can also be displayed at an adjacent position of the expression mark. The adjacent position can be above or below the expression mark, etc. By displaying the expression mark corresponding to the itinerary description information and the expression mark description information, the interface of the client can be made simple and the displayed information can be more interesting, which can attract the user to interact with the target virtual object.

[0100] In an alternative implementation, it is also possible to obtain itinerary details instruction information, and the itinerary details instruction information is used to indicate the itinerary details generation rule.

[0101] In this embodiment, the rules for generating itinerary details may include the word count rule, the number rule, and the content rule of itinerary details. For example, the itinerary detail instruction information is to generate 1 to 3 pieces of itinerary detail information corresponding to any itinerary description information according to the first object information, historical memory information, itinerary outline, and any itinerary description information. Each piece of itinerary detail information does not exceed 100 words. Then, the word count rule of itinerary details is not to exceed 100 words, the number rule of itinerary details is not less than 1 piece and not more than 3 pieces of itinerary detail information, and the content rule of itinerary details is the itinerary detail information related to the first object information, historical memory information, itinerary outline, and any itinerary description information.

[0102] Optionally, the itinerary detail instruction information may be input by the administrator corresponding to the server at a preset time. For example, the administrator inputs the itinerary detail instruction information at zero o'clock every day, so that the target large language model performs itinerary detail generation processing on the first object information, historical memory information, itinerary outline, and any itinerary description information according to the itinerary detail generation rules, and generates the itinerary detail information corresponding to any itinerary description information. The itinerary detail instruction information may also be automatically generated through a trigger mechanism. The trigger mechanism may be a preset trigger time. For example, when reaching zero o'clock every day, the server will automatically generate the itinerary detail instruction information.

[0103] Optionally, the outline instruction information, the itinerary description instruction information, and the itinerary detail instruction information may be input together by the administrator corresponding to the server. For example, the outline instruction information, the itinerary description instruction information, and the itinerary detail instruction information are input together. When the server needs to generate an itinerary outline, it obtains the outline instruction information. When the server needs to generate itinerary description information, it obtains the itinerary description instruction information. When the server needs to generate itinerary detail information, it obtains the itinerary detail instruction information. Optionally, the outline instruction information, the itinerary description instruction information, and the itinerary detail instruction information may be input separately by the administrator corresponding to the server. For example, when the server is required to generate an itinerary outline, the outline instruction information is input. When the server is required to generate itinerary description information, the itinerary description instruction information is input. When the server is required to generate itinerary detail information, the itinerary detail instruction information is input.

[0104] In an alternative embodiment, the target large language model may also perform itinerary detail generation processing on the first object information, historical memory information, itinerary outline, and any itinerary description information according to the itinerary detail generation rules, and generate the itinerary detail information corresponding to any itinerary description information.

[0105] In this embodiment, the target large language model generates itinerary details information corresponding to any itinerary description information by performing itinerary details generation processing on the first object information, historical memory information, itinerary outline, and any itinerary description information according to the itinerary details generation rules indicated by the itinerary details instruction information, such as the word count rule of the itinerary details, the number rule of the itinerary details, and the content rule of the itinerary details. For example, the first object information, historical memory information, itinerary outline, and any itinerary description information can be subjected to feature vector extraction through the embedding layer of the target large language model. The encoder layer understands the input feature vectors through the self-attention mechanism, and the decoder layer interprets the required information based on the input feature vectors and combines the self-attention and multi-head attention mechanisms, and finally gradually generates each element of the itinerary details information (such as the itinerary description content per minute). Optionally, the target large language model can also perform itinerary details generation processing on the first object information, second object information, historical memory information, itinerary outline, and any itinerary description information according to the itinerary details generation rules to generate itinerary details information corresponding to any itinerary description information. For the specific steps, please refer to the above description, and this solution will not be elaborated here. Among them, the first object information and the second object are input into the target large language model again to ensure that the itinerary details information generated by the large language model does not violate common sense. By effectively utilizing the native description and expression capabilities of the LLM, and expanding imagination based on certain conditions (the first object information, the second object information, historical memory information, itinerary outline, and any itinerary description information), scenarios, content, inner monologues, etc. of the itinerary details information of the style type suitable for the current user's interests are constructed.

[0106] In an alternative embodiment, it is also possible to obtain the fourth object information, preset historical memory information, preset itinerary outline, preset itinerary description information, preset itinerary details information, and itinerary details instruction information for training the virtual object; the itinerary details instruction information is used to indicate the itinerary details generation rules; the initial large language model performs itinerary details generation processing on the fourth object information, preset historical memory information, preset itinerary outline, and preset itinerary description information according to the itinerary details generation rules to generate itinerary details information corresponding to the preset itinerary description information; aiming to reduce the difference between the itinerary details information corresponding to the preset itinerary description information and the preset itinerary details information, the initial large language model is trained, and after the training is completed, the target large language model is obtained; the target large language model is used to perform itinerary details generation processing on the first object information, historical memory information, itinerary outline, and any itinerary description information according to the itinerary details generation rules to generate itinerary details information corresponding to any itinerary description information. By training the target large language model through the above method, the target large language model can accurately generate itinerary details information according to the first object information, historical memory information, itinerary outline, and any itinerary description information.

[0107] In an alternative embodiment, world rule information may also be obtained; the target large language model performs itinerary details generation processing on the first object information, historical memory information, itinerary outline, any itinerary description information, and world rule information according to the itinerary details generation rules indicated by the itinerary details instruction information to generate itinerary details information corresponding to any itinerary description information.

[0108] In an alternative embodiment, the initial itinerary time interval corresponding to any itinerary description information may be obtained; based on the preset time constraint, the initial itinerary time interval is processed to be narrowed down to obtain the target itinerary time interval; the target large language model performs itinerary details generation processing on the first object information, historical memory information, itinerary outline, target itinerary time interval, and any itinerary description information according to the itinerary details generation rules to generate itinerary details information corresponding to any itinerary description information; the itinerary content corresponding to the itinerary details information is carried out within the target itinerary time interval.

[0109] In this embodiment, for example, the initial itinerary time interval is from 9 o'clock to 10 o'clock, and the preset time constraint is 5 to 10 minutes. Processing to narrow down the initial itinerary time interval may be randomly selecting 6 minutes from the preset time constraint of 5 to 10 minutes, and using this 6 minutes to process the initial itinerary time interval to obtain the target itinerary time interval from 9:06 to 9:54. Then, the target large language model performs itinerary details generation processing on the first object information, historical memory information, itinerary outline, target itinerary time interval, and any itinerary description information according to the itinerary details generation rules to generate itinerary details information corresponding to any itinerary description information, which can ensure that the generated itinerary details information is all carried out within the target itinerary time interval, effectively reducing phenomena such as itinerary time jumps and disorders.

[0110] In an alternative embodiment, the itinerary details instruction information includes an itinerary details example corresponding to the itinerary description information, and the itinerary details example includes the target corresponding to the itinerary description information. Among them, the itinerary details example including the target corresponding to the itinerary description information can be understood as the generated itinerary details information meeting the itinerary details example.

[0111] Among them, a normal form can be used as an example of the itinerary details information generated by the target large language model each time. The normal form used is a dynamic normal form, that is, each time the details information is generated, the time interval corresponding to the normal form will be updated to the target itinerary time interval corresponding to any itinerary description information. Optionally, the time corresponding to the first normal form can be set as the starting time point of the target itinerary time interval, for example, 9:06, and the time of the second normal form can be set as the ending time point of the target itinerary time interval, for example, 9:54. When the target large language model generates itinerary details information, by combining the first normal form, the time corresponding to the first normal form, the second normal form, and the time corresponding to the second normal form, the target large language model can better understand the target itinerary time interval, and ensure that the itinerary content corresponding to the generated itinerary details information is within the target itinerary time interval, effectively reducing phenomena such as itinerary time jumps and disorders.

[0112] Optionally, if the target scenario type is a memory scenario, then the target large language model can perform itinerary details generation processing on the first object information, second object information, third object information, historical memory information, itinerary outline, and any itinerary description information according to the itinerary details generation rules, and generate the itinerary details information corresponding to any itinerary description information. Combining historical memory information when generating itinerary details information can enhance the relevance between the target virtual object and the target role object (user), and improve the user's immersion and usage experience. The itinerary details instruction information corresponding to different target scenario types is different. The itinerary details instruction information for the memory scenario can be I 3m , then the itinerary details information for the memory scenario can be expressed by formula (7) as:

[0113] T 3mi =φ(I 3m ,T 1m ,T 2mi ,P p ,P u ,N,M),i∈{1,2,…,K3} (7)

[0114] Among them, T 3mi represents the i-th itinerary details information generated, I 3m represents the itinerary details instruction information for the memory scenario, T 1m represents the itinerary outline for the memory scenario, T 2mi represents the itinerary description information for the memory scenario, P p represents the first object information, P u represents the second object information, N represents the third object information, and M represents the historical memory information.

[0115] Optionally, if the target scenario type is a non-memory scenario, the target large language model processes the first object information, the second object information, the third object information, the itinerary outline, and any itinerary description information according to the itinerary details generation rule to generate itinerary details information corresponding to the itinerary description information. The non-memory scenario can include a normal scenario and an abnormal scenario. When there is no interaction between the target role object and the target virtual object within the preset interaction time, it can be determined that the target scenario type of the target virtual object is an abnormal scenario. The itinerary details instruction information for the normal scenario can be I 3n , then the itinerary details information for the normal scenario can be expressed by formula (8) as:

[0116] T 3ni = φ(I 3n , T 1n , T 2ni , P p , P u , N), i ∈ {1, 2, …, K1} (8)

[0117] Among them, T 3ni represents the i-th itinerary details information generated, I 3n represents the itinerary details instruction information for the normal scenario, T 1n represents the itinerary outline for the normal scenario, T 2ni represents the itinerary description information for the normal scenario, P p represents the first object information, P u represents the second object information, and N represents the third object information.

[0118] The itinerary details instruction information for the abnormal scenario can be I 3a , then the itinerary details information for the abnormal scenario can be expressed by formula (9) as:

[0119] T 3ai = φ(I 3a , T 1a , T 2ai , P p , P u , N), i ∈ {1, 2, …, K2} (9)

[0120] Among them, T 3ai represents the i-th itinerary details information generated, I 3a represents the itinerary details instruction information for the abnormal scenario, T 1a represents the itinerary outline for the abnormal scenario, T 2ai represents the itinerary description information for the abnormal scenario, P p represents the first object information, P u represents the second object information, and N represents the third object information.

[0121] Each generation of a travel detail information is a call to the target large language model once.

[0122] By designing different scenarios to generate the corresponding travel detail information, the generated travel detail information can be more in line with the user's interest and willingness, attracting the user to interact with the target virtual object, ensuring the user experience, and enhancing user stickiness.

[0123] In the above implementation, by decomposing the daily itinerary of the target virtual object into multiple levels, the advantages of the target large language model itself in reading comprehension, logical planning, and language expression are effectively utilized to generate the itinerary outline, itinerary description information, and itinerary detail information of the target virtual object respectively. Finally, the generation of complete itinerary content is realized, and the simulation of the behavior of a complete target virtual object is achieved, thus laying a complete target virtual object image for the user's emotional companionship process.

[0124] S205, Send at least one itinerary description information to the client corresponding to the target role object, and the client is used to display at least one itinerary description information.

[0125] In this embodiment, the client corresponding to the target role object can be determined through the object identifier of the target role object, and at least one itinerary description information is sent to the client corresponding to the target role object. Among them, the object identifier of the target role object can be the role account of the target role.

[0126] Optionally, at least one itinerary description information within a preset time period can be displayed through the itinerary description interface of the client.

[0127] Optionally, at least one piece of trip description information is arranged and displayed in the trip description interface in any of the following ways: the trip time corresponding to the trip description information, the trip type corresponding to the trip content, the area type corresponding to the activity area, and the emotion type of the target virtual object when performing the corresponding activity. For example, the trip description information for describing each trip content can be arranged and displayed in the trip description interface in the order of the trip time corresponding to the trip description information from morning to night. Another example is that the trip description information for describing the trip content of the same trip type can be displayed in a display area of the trip description interface, and the trip description information for describing the trip content of different trip types can be displayed in different display areas of the trip description interface. The trip types can include feeding, walking, going to school, resting, etc. Another example is that the trip description information for describing the trip content generated by activities in the same activity area can be displayed in a display area of the trip description interface, and the trip description information for describing the trip content generated by activities in different activity areas can be displayed in different display areas of the trip description interface. The activity areas can include schools, parks, homes, etc. Another example is that the trip description information of the target virtual object when in the same emotion type during activities can be displayed in a display area of the trip description interface, and the trip description information of the target virtual object when in different emotion types during activities can be displayed in different display areas of the trip description interface. The emotion types can include happy, frustrated, angry, frightened, etc.

[0128] Taking Figure 3A as an example, Figure 3A is a schematic diagram of the trip description interface provided by an embodiment of the present application. At least one piece of trip description information in this trip description interface can be arranged and displayed according to the trip time corresponding to the trip description information. For example, assuming that the preset time period is a time period less than or equal to three days from the current system time, then the trip content generated by the target virtual object's activities in the virtual scene within the last three days can be displayed in the trip description interface.

[0129] Optionally, at least one piece of trip description information for the Nth day in the preset time period can be displayed. Any piece of trip description information in the at least one piece of trip description information can correspond to a trip node. Each piece of trip description information in the at least one piece of trip description information is used to describe the trip content of the target virtual object at this trip node, and N is a positive integer. Each piece of trip description information can include the trip time and the trip content. Each trip node can correspondingly display the trip time of this trip node. Taking Figure 3AFor example, the Nth day can be the most recent day, i.e., today. The first itinerary node today can be "make coffee at home", and the itinerary time of the first itinerary node can be 15:00, indicating that the target virtual object made coffee at 15:00. At the itinerary node corresponding to the first itinerary node, three itinerary details are generated, and the activity content described in these three itinerary details is associated with the itinerary node. For example, the second itinerary detail corresponding to the first itinerary node is "Xiaopa selected a bag of coffee beans that were just bought recently. Xiaoling said they are the top 10 beans in the world and are limited edition!". The generation time corresponding to the second itinerary detail can be 15:20, indicating that the target virtual object Xiaopa selected a bag of coffee beans that were just bought recently at 15:20. Optionally, if the number of itinerary details corresponding to any itinerary node is greater than the first preset quantity threshold, then the itinerary details with the quantity of the first preset quantity threshold and the first remaining quantity prompt message can be displayed. For example, Figure 3A Among them, the itinerary node "surprise! dinner blind box" corresponding to the time information today has five itinerary details. Assuming that the first preset quantity threshold is 2, then the client can display the first two itinerary details below this time information today, and display the first remaining quantity prompt message "Expand 3 records" below the last itinerary detail, indicating that for the itinerary node "surprise! dinner blind box" corresponding to the time information today, in addition to the two displayed itinerary details, there are also three itinerary details. Optionally, for any itinerary node, if the number of corresponding itinerary details is less than or equal to the first preset quantity threshold, the client can display each itinerary detail corresponding to this itinerary node. Optionally, the client can display each itinerary detail corresponding to this itinerary node in response to the trigger operation on the first remaining quantity prompt message.

[0130] Optionally, at least one itinerary node description information of the target virtual object in the virtual environment generated by the target large language model for each day except the Nth day in the preset time period can be displayed in the itinerary description interface, and the itinerary nodes of the same day can be displayed in the same area of the itinerary description interface. For example, if three itinerary nodes corresponding to the target virtual object were generated yesterday by the target large language model, then the time information, i.e., "yesterday", can be displayed in the itinerary description interface, and below the time information, the three itinerary nodes corresponding to the target virtual object yesterday can be displayed, such as "being scolded by the teacher for being stupid in today's English class", "going to the playground for a run to relax", and "a football game with friends". Optionally, an expansion control can be displayed at the associated position of each itinerary node, and the client can display at least one itinerary detail corresponding to this itinerary node below the itinerary node corresponding to the trigger operation on any expansion control.

[0131] Optionally, the itinerary description information before the preset time period can be hidden in the itinerary description interface. For example, only the time information of each day before the preset time period and the emoji icons corresponding to at least one itinerary node are displayed. The emoji icons are used to represent the state of the virtual object at the corresponding itinerary node. Optionally, for any time information, if the number of corresponding emoji icons is greater than the second preset quantity threshold, the client can display the emoji icons with the quantity of the second preset quantity threshold and the second remaining quantity prompt information. For example, Figure 3A in Figure 3A , the time information of November 29 corresponds to five itinerary nodes. Assuming that the second preset quantity threshold is 3, then the client can display the emoji icons corresponding to three of the five itinerary nodes on the right side of the time information of November 29, and display the second remaining quantity prompt information "+2" on the right side of the last emoji icon, indicating that in addition to the itinerary nodes corresponding to the three displayed emoji icons for the time information of November 29, there are also two itinerary nodes. Optionally, for any time information, if the number of corresponding emoji icons is less than or equal to the second preset quantity threshold, the client can display the emoji icons corresponding to each itinerary node of the time information. Taking Figure 3B as an example, Figure 3B FIG. Figure 3B is a schematic diagram of the itinerary description interface provided by an embodiment of the present application. The client can display each itinerary node corresponding to the time information in response to a trigger operation on any time information or the area where the corresponding emoji icon is located. The associated position of each itinerary node can display a corresponding expansion control. The client can display at least one itinerary detail information corresponding to the itinerary node below the expansion control corresponding to the itinerary node in response to a trigger operation on any expansion control. If the number of itinerary detail information corresponding to a certain itinerary node is greater than the first preset quantity threshold, then the itinerary detail information with the quantity of the first preset quantity threshold and the first remaining quantity prompt information can be displayed. The client can display each itinerary detail information corresponding to the itinerary node in response to a trigger operation on the first remaining quantity prompt information.

[0132] Optionally, the associated position of any itinerary node in the itinerary description interface can correspondingly display the emoji icon corresponding to the itinerary node. Taking Figure 3A as an example, the emoji icon can be displayed on the left side of the itinerary node. Optionally, the emoji icon can be displayed above the itinerary time of the itinerary node.

[0133] Among them, the emoji icon can include the emoji icon. The emoji icon can be used to represent the emotion or activity of the target virtual object at the corresponding itinerary node. Taking Figure 3AFor example, if the target virtual object was happy during a football game with friends at 14:30 yesterday, the expression identifier corresponding to the itinerary node "Football game with friends" can be a smiling face emoji (happy mood). If the target virtual object went for a run on the playground at 14:10 yesterday, the expression identifier corresponding to the itinerary node "Go for a run on the playground to relax" can be a shoe pattern (running activity).

[0134] Optionally, the time information can be a first-level heading, the itinerary node (itinerary description information) can be a second-level heading, and the itinerary details information can be a third-level heading.

[0135] In this embodiment, the itinerary description information describes the virtual world of the target virtual object. The interaction between the user and the target virtual object can be in the real world, such as answering questions raised by the user or making some actions to respond to the user, etc. Therefore, this embodiment can not only increase the fun of the interaction between the user and the target virtual object, but also achieve the information interoperability between the virtual world and the real world, creating a more real, interesting, and interactive virtual object interaction experience.

[0136] In an alternative embodiment, the ways for the client to display the itinerary description interface can include the following two:

[0137] 1. The client can display the itinerary track interface of the target virtual object. The itinerary track interface includes the itinerary track information of the target virtual object within a preset time period. The itinerary track information includes at least one itinerary node. One piece of itinerary description information corresponds to one itinerary node, and each itinerary node corresponds to one or more pieces of itinerary details information. The client can also, in response to a trigger operation on any one of the at least one itinerary node, display the itinerary description interface and display one or more pieces of itinerary details information corresponding to the itinerary node in the itinerary description interface. Among them, the itinerary details information corresponding to any one itinerary node is used to describe the itinerary details content of the target virtual object at the itinerary node indicated by the itinerary node.

[0138] For Figure 3C example,[[]] Figure 3C is a schematic diagram of the itinerary track interface provided by the embodiment of the present application. Each itinerary node in the itinerary track interface can be arranged and displayed in the order of the corresponding generation time from early to late. The last itinerary node can be connected to the target virtual object to represent the arrangement order of each itinerary node. Optionally, the itinerary nodes in the itinerary track interface can be represented by the corresponding itinerary time and expression identifier. If the user wants to view the itinerary details information of a certain itinerary node, the user can click on the itinerary node corresponding to the itinerary node. The client can, in response to the trigger operation on the itinerary node, display Figure 3CThe travel description interface is displayed, and the travel description information corresponding to the travel node and one or more pieces of travel detail information corresponding to the travel description information are displayed in the travel description interface. The user can view the travel detail information corresponding to other travel nodes by swiping up and down the travel description interface.

[0139] Second, the client can display the travel track interface of the target virtual object, and the travel track interface includes the status information of the target virtual object. The client can also display the travel description interface in response to a trigger operation on the status information.

[0140] Take Figure 3D as an example. Figure 3D FIG. is a schematic diagram of the travel track interface provided by an embodiment of the present application. The status information of the target virtual object can be displayed above the travel track information in the travel track interface. The status information is used to describe the activity content or highlight moment of the target virtual object at the target travel node. The target travel node can be any travel node, the travel node corresponding to the highlight moment, or the travel node with the smallest interval from the current system time. Optionally, the client can display a part of the status information of the virtual object in the travel track interface of the target virtual object. The user can display the complete content of the status information of the target virtual object by clicking on the area where the status information is located and swiping down. If the user wants to know the specific travel of the target virtual object based on the complete content of the status information of the target virtual object, then the user can click on the status information, and the client can display the travel description interface in response to the trigger operation on the status information.

[0141] In one embodiment, the travel track interface may include at least one of the following: the status information of the target virtual object, the travel track information of the target virtual object within the preset time period, the target virtual object, the session message input area, the background image of the travel track interface;

[0142] Among them, the status information includes at least one of the following: the travel time of the target travel node, the expression identifier corresponding to the travel description information, the description information of the expression identifier, and the travel content corresponding to the target travel node.

[0143] Among them, the travel trajectory information includes at least one travel node, and the at least one travel node is arranged and displayed according to the corresponding travel time. The sizes of the expression icons of these travel nodes arranged in sequence can increase or remain unchanged along with the sequence, and the amount of information displayed can also increase or remain unchanged along with the sequence. For example, the travel node information of the travel node with the longest time interval from the current system time in the travel trajectory information may include an expression icon; the travel node information of the travel node with a relatively long time interval from the current system time may include an expression icon and the travel time of the travel node; the travel node information of the travel node with a relatively short time interval from the current system time may include an expression icon, the travel time of the travel node, and expression icon description information. The expression icon description information is, for example, a location such as a beach, a park, a quilt, etc., or it can also be an emotion; the travel node information of the travel node with the shortest time interval from the current system time may include an expression icon, the travel time of the travel node, expression icon description information, and travel content. The travel content is, for example, "Xiaopa chased butterflies 40 times in the park" or "rolled in a lot of sand", etc.

[0144] Optionally, the travel trajectory information may further include unknown travel nodes, that is, travel nodes to be unlocked, which are the travel nodes that the target virtual object is about to generate after the target travel node. The unknown travel nodes can be arranged and displayed before the target travel node, and the unknown travel nodes are represented by preset expression icons to indicate that the travel node is an unknown travel node.

[0145] Among them, the background image can be generated by a text-to-image model based on the status description of the target travel node and / or the current environmental information. The environmental information may include weather information, time information, temperature information, or ambient light brightness information, etc., of the environment where the user is currently located. The environmental information of the environment where the user is currently located can be collected by the computer device running the client. For example, different background images correspond to day and night respectively, and different background images correspond to rainy days, sunny days, or cloudy days respectively.

[0146] Optionally, the client can display a travel trajectory interface. The travel trajectory interface may include the status information of the target virtual object. The status information may include at least one of the following information: the travel time of the target travel node, the expression icon corresponding to the travel description information, the expression icon description information, and the travel content corresponding to the target travel node. For example, the expression icon can be an emoji, such as an emoticon or an object symbol, etc. The expression icon description information may include a location or an emotion, etc. Figure 3DFor example, the status information can be located in the top area of the travel track interface. For example, the travel content corresponding to the target travel node can be "Xiaopa chased butterflies 40 times in the park", the expression identifier can be a butterfly pattern, the travel time can be "19:30", and the description information of the expression identifier can be "chasing butterflies". Optionally, the character length of the travel content displayed in the travel track interface does not exceed the third preset quantity threshold. For example, assuming the third preset quantity threshold is k, where k is a positive integer, the first k characters of the travel content can be displayed in the top area of the travel track interface. Another example is that when the character length of the travel content exceeds the third preset quantity threshold, the travel content can be scrolled and displayed in the top area of the travel track interface.

[0147] Optionally, according to the mapping relationship between the travel description information and the expression identifier, determine the expression identifier corresponding to any travel description information. Optionally, the character length of the expression identifier does not exceed the fourth preset quantity threshold. Specifically, the travel description information can include time, location, person, person relationship, mood, event, environmental description, etc. The pre-created expression identifier database can include multiple expression identifiers, and the expression identifiers can be classified into categories such as mood, food, animal, plant, sports, building, tool, expression, etc. Different categories correspond to different priorities. If multiple expression identifiers are matched based on the travel summary, then the expression identifier with the highest priority can be selected from the expression identifiers matched by the travel description information, and the selected expression identifier can be used as the matched emoji. Optionally, if no expression identifier matching the travel description information is found, then a candidate expression identifier can be randomly selected from the candidate expression identifier cluster as the emoji obtained by matching the travel description information. The candidate expression identifier cluster can include at least one candidate expression identifier, that is, a fallback expression identifier. Optionally, the expression identifier matching the travel description information can be found by means of exact matching or fuzzy matching, and the fuzzy matching can include near-synonym relationship matching, etc.

[0148] For example, if the itinerary content included in the itinerary description information expresses an emotion, the expression identifier may include an emoji that matches the emotion, and the expression identifier description information may include emotion information for describing the emotion. If the itinerary summary expresses a behavioral attitude, the expression identifier may include an emoji that matches the behavioral attitude, and the expression identifier description information may include behavior information for describing the behavioral attitude. If the itinerary content included in the itinerary description information expresses a sports event, the expression identifier may include an emoji that matches the sports event, and the expression identifier description information may include sports information for describing the sports event. If the itinerary content included in the itinerary description information expresses an activity event, the expression identifier may include an emoji that matches the activity event, and the expression identifier description information may include activity information for describing the activity event. If the itinerary content included in the itinerary description information expresses food, the expression identifier may include an emoji that matches the food, and the expression identifier description information may include the name of the food.

[0149] Optionally, the client can also play background music when displaying the itinerary track interface, and the background music can be set to on or off. For example, the background music can be defaulted to off, and when the user wants to play the background music, the user can set it to on. Another example is that the background music can be defaulted to on, and when the user wants to pause playing the background music, the user can set it to off.

[0150] In an optional implementation manner, after the client displays the itinerary track interface, if no interaction operation is received within the third preset duration, at least one itinerary node displayed in the itinerary track interface can be updated, so as to display the updated itinerary node in the itinerary track interface.

[0151] In an optional implementation manner, itinerary update prompt information and some or all of the updated itinerary description information can be displayed in the itinerary description interface. Among them, the itinerary update prompt information is used to prompt that there is at least one updated itinerary description information

[0152] Take Figure 3E as an example Figure 3EIt is a schematic diagram of the itinerary track interface provided by an embodiment of the present application. If the user has not browsed the itinerary description interface for a period of time, then there will be at least one updated itinerary description information in the itinerary description interface, that is, the itinerary description information that the user has not read. The itinerary description information that was most recently generated can also be displayed in the itinerary description interface. Then, when the itinerary description interface is displayed on the client, itinerary update prompt information can be displayed in the itinerary description interface. For example, "17 updates" indicates that the user has 17 unread itinerary description information. Optionally, an expression icon with a quantity less than or equal to a third quantity threshold can also be displayed at an associated position of the itinerary update prompt information. The expression icon is the expression icon corresponding to the itinerary node to which the updated itinerary description information belongs. The itinerary update prompt information can be used to prompt the existence of unread itinerary description information. The client can display the M earliest generated updated itinerary description information in response to a trigger operation on the itinerary update prompt information. When displaying the M earliest generated updated itinerary description information, a positioning control can also be displayed. The client can display the most recently generated itinerary description information in response to a trigger operation on the positioning control, where M is a positive integer. The generation time of the most recently generated itinerary description information is later than the generation time of the M updated itinerary description information.

[0153] In an alternative embodiment, the itinerary description interface may include a back button. The client can display the itinerary track interface of the target virtual object and display the target virtual object in the itinerary track interface in response to a trigger operation on the back button.

[0154] Among them, the preset information of the target virtual object displayed in the itinerary track interface is related to the interaction with the target virtual object. The preset information includes at least one of the following: posture information, expression information, behavior information, itinerary planning information of the target virtual object within a third preset time period, and preference information of the virtual object.

[0155] Take Figure 3D as an example. The back button can be located in the top area of the itinerary description interface. The top area may include a description interface identifier (such as "Xiaopa's World") and the back button. The back button is located to the left of the description interface identifier, and the user can display the itinerary track interface by clicking on the back button. Optionally, the back button can be located in the bottom area of the itinerary description interface. The bottom area may include the back button, and the text information "Return to Home Page" is displayed on the back button. The user can display the itinerary track interface by clicking on the back button.

[0156] In this embodiment, at least one of the pose information, expression information, behavior information of the target virtual object displayed in the travel trajectory interface, the travel plan information of the target virtual object within the third preset time period, and the preference information of the target virtual object is related to the interaction with the target virtual object. By endowing the user and the target virtual object with more interesting and personalized experiences, and empowering with Artificial Intelligence (AI) technology, more levels of interactivity can be brought to the virtual object application, enabling the virtual object to exhibit more intelligent behaviors and personalized features, and being closer to the identity of the user's "friend", thereby shaping a more interesting, intimate and vivid virtual partner for the player, and establishing a deeper emotional bond between the user and the virtual object.

[0157] In the embodiment of the present application, the first object information of the target virtual object, the second object information of the target role object, and the outline instruction information are obtained; the outline instruction information is used to indicate the outline generation rule; the target large language model performs travel outline generation processing on the first object information and the second object information according to the outline generation rule to generate a travel outline of the target virtual object within the preset time period; the travel description instruction information is obtained, and the travel description instruction information is used to indicate the travel description generation rule; the target large language model performs travel description generation processing on the first object information and the travel outline according to the travel description generation rule to generate at least one travel description information corresponding to the travel outline; the at least one travel description information is sent to the client corresponding to the target role object, and the client is used to display the at least one travel description information. In the embodiment of the present application, the travel outline and the travel description information are generated hierarchically, the travel content is gradually refined, and combined with the generation ability of the large language model for different tasks, the travel description information of the target virtual object with text content can be generated at a lower cost, realizing the lightweight generation of the interactive content of the virtual object. In addition, even when the user does not initiate an interaction instruction, the virtual pet will actively generate corresponding interactive content, which can attract the user to interact with the virtual object to enhance user viscosity.

[0158] Based on the above description, please refer to Figure 4 , Figure 4 which is a schematic flowchart of another virtual object interaction method provided by the embodiment of the present application. As Figure 4 shown, the virtual object interaction method includes but is not limited to steps S401-S411, where:

[0159] S401. The server obtains the first object information of the target virtual object, the second object information of the target role object, and the outline instruction information; the outline instruction information is used to indicate the outline generation rule.

[0160] For the specific implementation steps of this solution, please refer to the above step S201, and this solution will not be elaborated here.

[0161] S402: The server performs itinerary outline generation processing on the first object information and the second object information according to outline generation rules using the target large language model to generate a itinerary outline for the target virtual object within a preset time period.

[0162] For the specific implementation steps of this solution, please refer to the above step S202, which will not be repeated in this solution.

[0163] S403: The server obtains itinerary description instruction information, where the itinerary description instruction information is used to indicate itinerary description generation rules.

[0164] The specific implementation steps of this solution can be found in the above step S203, which will not be described in detail in this solution.

[0165] S404: The server performs a trip description generation process on the first object information and the trip outline according to the trip description generation rule using the target large language model to generate at least one piece of trip description information corresponding to the trip outline.

[0166] The specific implementation steps of this solution can be found in the above step S204, which will not be repeated in this solution.

[0167] S405: The server sends at least one piece of itinerary description information to the client.

[0168] S406: The client displays at least one itinerary description information.

[0169] The specific implementation steps of this solution can be found in the above step S205, which will not be described in detail in this solution.

[0170] S407: The client sends a session message to the server.

[0171] In this embodiment, the user can send a conversation message to the target virtual object through the conversation message input area, and the client sends the conversation message to the server so that the server generates a reply message. The conversation message input area can be present in the travel track interface.

[0172] Optionally, the travel track interface further includes a greeting message sent by the target virtual object. The client can display a conversation interface in response to clicking on the greeting message. The greeting message is used to guide the target role object (user) to interact with the target virtual role. Optionally, after the client displays the greeting message in the travel track interface, if no interaction operation is received within the first preset duration, then the client can cancel the display of the target virtual object and the greeting message in the travel track interface. Further optionally, the client can display the target virtual object and the greeting message again in the travel track interface every second preset duration. The target virtual object and the greeting message displayed each time are related to the current displayed status information. Therefore, the target virtual object and the greeting message displayed at different times may be different. Optionally, a quick reply control can be displayed at the associated position of the greeting message. When the user clicks on the quick reply control, a conversation message is sent in the conversation interface of the target virtual object. The client can keep the travel track interface displayed unchanged and display a send prompt message "Message sent", and then the quick reply control is cancelled.

[0173] Optionally, when the target travel node is updated, the status information displayed in the top area of the travel track interface is updated. Since the target virtual object and the greeting message displayed in the travel track interface are related to the current displayed status information, when the target travel node is updated, the target virtual object and the greeting message displayed in the travel track interface are also updated. For example, when the target travel node is the travel node with the smallest time interval from the current system time, if a new travel node is detected, then the target travel node is updated, the status information displayed in the top area of the travel track interface is updated, and the target virtual object and the greeting message displayed in the travel track interface are also updated.

[0174] Optionally, the client can display a conversation interface in response to clicking on the greeting message, clicking on the conversation message input area, or pulling up the travel track interface. Optionally, the travel track interface can further include a conversation message input area, which can include a conversation message input prompt "Say something to Xiaopa...". The user can send a conversation message to the target virtual object through the conversation message input area.

[0175] Optionally, an interaction interface with the target virtual object can be displayed in response to a trigger operation on any travel detail information.

[0176] In an optional implementation, the interaction interface can include a conversation interface. Then the way for the client to interact with the target virtual object in the interaction interface can include: outputting the conversation message input in the conversation message input area to the target virtual object in response to an input operation on the conversation message input area in the conversation interface.

[0177] toFigure 3F For example, Figure 3F FIG. Figure 3F is a schematic diagram of a travel description interface provided by an embodiment of the present application. The client can display a session interface with a target virtual object in response to a trigger operation on any travel detail information. The session interface may include a session message input area. The client can output the session message input in the session message input area to the target virtual object in response to an input operation on the session message input area in the session interface. Optionally, the user can input text information, voice information, image information, video information, etc. in the session message input area.

[0178] Optionally, the client can also display reference information in the session message input area of the session interface. The reference information has a reference relationship with the session message input by the input operation. Optionally, the client can display the session message sent in the identity of the target role object in the session interface, as well as the reference information referred to by the session message.

[0179] Among them, the reference information includes at least one of the following: any travel detail information, an element associated with any travel detail information, interaction prompt information, and an element associated with the interaction prompt information. The element associated with any travel detail information may include keywords in the travel detail information, a topic corresponding to the travel detail information, or other virtual objects or role objects mentioned in the travel detail information. The interaction prompt information may include an object identifier of a virtual object mentioned in any travel detail information, such as an object pattern or an object name, etc.

[0180] Optionally, after the user inputs a session message through the session message input area, the client can send the session message in the identity of the target role object in the session interface and display the reference relationship between the session message and the reference information. This reference relationship can facilitate the target virtual object to determine the user's interaction motivation, and then determine a reply message based on the reference information and the session message.

[0181] Take Figure 3F For example, if the user wants to interact with the target virtual object "Xiaopa" when browsing the travel detail information "Xiaopa had no idea, so he ordered a blind box takeaway - unexpectedly, it tasted great", it indicates that the travel detail information "Xiaopa had no idea, so he ordered a blind box takeaway - unexpectedly, it tasted great" has inspired the user's desire to interact with the detail virtual object. Therefore, reference information can be displayed in the session message input area of the session interface to prompt that the motivation for the user to interact with the detail virtual object is the travel detail information "Xiaopa had no idea, so he ordered a blind box takeaway - unexpectedly, it tasted great".

[0182] Take Figure 3G For example, Figure 3GIt is a schematic diagram of the itinerary description interface provided by an embodiment of the present application. If a certain itinerary detail information in the itinerary description interface mentions other virtual objects, such as "Xiaoling", then interactive prompt information can be displayed below this itinerary detail information. The interactive prompt information can be, for example, "Go and chat with the new friend Xiaoling". If the user wants to interact with the virtual object "Xiaopa" when browsing this itinerary detail information, that is, it indicates that this itinerary detail information stimulates the user's desire to interact with the target virtual object, then the user can click on the interactive prompt information corresponding to this itinerary detail information. Therefore, reference information can be displayed in the session message input area of the session interface. The reference information can include the object identifier in the interactive prompt information to prompt the user that the topic for interacting with the target virtual object is the new friend Xiaoling.

[0183] In an alternative implementation, the triggering operation on any itinerary detail information can include the following two methods:

[0184] 1. In response to the triggering operation on any itinerary detail information, display an interactive control, and in response to the triggering operation on the interactive control, display an interactive interface.

[0185] As Figure 3F shown, the user can click on any itinerary detail information. The client can, in response to the triggering operation on any itinerary detail information, display an interactive control above this itinerary detail information. After the user clicks on this interactive control, the client can, in response to the triggering operation on the interactive control, display an interactive interface. The interactive control displays the text information "Go and chat".

[0186] 2. When interactive prompt information is displayed at the associated position of any itinerary detail information in the itinerary description interface, in response to the triggering operation on the interactive prompt information, display an interactive interface.

[0187] As Figure 3G shown, if a certain itinerary detail information in the trajectory description interface mentions other virtual objects, such as "Xiaoling", then interactive prompt information can be displayed below this itinerary detail information. The interactive prompt information can be, for example, "Go and chat with the new friend Xiaoling". The user can click on this interactive prompt information. The client can, in response to the triggering operation on the interactive prompt information, display an interactive interface.

[0188] Optionally, when switching from the interactive interface to the itinerary trajectory interface, mood information can be sent in the itinerary trajectory interface in the identity of the target virtual object, and the mood described in the mood information is related to the interaction process of the interactive interface.

[0189] S408. The server obtains session instruction information, historical memory information, and context information. The session instruction information is used to indicate the reply message generation rule.

[0190] In this embodiment, the context information is used to indicate the session messages between the target virtual object and the target role object before the sending time of the target session message, and the session messages sent by the target role object after the sending time of the target session message. The time difference between the sending time of the target session message and the current system time is less than or equal to the sending time of other session messages sent by the target virtual object. The target session message is sent in the identity of the target virtual object. Among them, the context information may include the session message newly sent by the client and the session messages between the target virtual object and the target role object within a preset duration before the newly sent session message; or, the context information may include the session message newly sent by the client and the session messages of a preset number of messages between the target virtual object and the target role object before the newly sent session message. The reply message generation rule at least includes a message content rule. Optionally, the session instruction information may be input by the administrator after the server receives the session message sent by the client; or, the session instruction information may be automatically generated by the server after the server receives the session message sent by the client.

[0191] S409. The server performs reply message generation processing on the first object information, the second object information, the context information, the historical memory information, and the itinerary outline according to the reply message generation rule through the target large language model to generate a reply message of the target virtual object.

[0192] In this embodiment, the first object information, the second object information, the context information, the historical memory information, and the itinerary outline can be subjected to feature vector extraction through the embedding layer of the target large language model. The encoder layer uses the self-attention mechanism to understand the input feature vectors. The decoder layer interprets the required information according to the input feature vectors and combines the self-attention and multi-head attention mechanisms, and finally generates a reply message of the target virtual object. Among them, the historical memory information is input into the target large language model as supplementary information of the target virtual object to assist the target large language model to better generate replies closer to the user, enhance the personalized experience of the user corresponding to the target virtual object, and enhance the immersion of the user. In addition, inputting the first object information, the second object information, the context information, and the itinerary outline into the target large language model can give the target large language model sufficient target virtual object information, the itinerary information of the target virtual object, the target role information, and the context information, so that the target large language model can reply to the user's relevant questions well in the image of a virtual object existing in the virtual world, so as to achieve an echo with the user's history, virtual object behavior, etc. in multiple aspects and build a more immersive experience.

[0193] Optionally, the server can also generate a reply message for the target virtual object by performing reply message generation processing on the first object information, second object information, context information, historical memory information, itinerary outline, and itinerary details information according to the reply message generation rule of the target large language model. For the specific steps of this solution, please refer to the above steps, and this solution will not be elaborated here.

[0194] In an alternative embodiment, it is also possible to obtain the fourth object information for training the virtual object, the fifth object information for training the role object, the preset context information, the preset historical memory information, the preset itinerary outline, the preset reply message, and the session instruction information; the session instruction information is used to indicate the reply message generation rule; the initial large language model performs reply message generation processing on the fourth object information, fifth object information, preset context information, preset historical memory information, and preset itinerary outline according to the reply message generation rule indicated by the session instruction information to generate a reply message for the training virtual object; aiming at reducing the difference between the reply message of the training virtual object and the preset reply message, the initial large language model is trained, and after the training is completed, the target large language model is obtained; the target large language model is used to perform reply message generation processing on the first object information, second object information, context information, historical memory information, and itinerary outline according to the reply message generation rule to generate a reply message for the target virtual object. By training the target large language model through the above method, the target large language model can accurately generate reply messages.

[0195] In an alternative embodiment, the reply message generation rule at least includes a message content rule; the message content rule is used to restrict the content of the session message input by the user to prevent the target large language model from generating illegal reply messages. If the context information does not meet the message content rule, the target large language model performs reply message generation processing on the context information according to the reply message generation rule indicated by the session instruction information to generate a reply message for the virtual object, and the reply message is used to guide the role object to output a session message that meets the message content rule. Among them, adding the message content rule to the prompt input to the target large language model enables the target large language model to generate safe, reliable, and user-expected conversation content. For example, if the user inputs some illegal conversations, sensitive information, or other illegal session messages, before generating a reply message, the target large language model will detect the session message input by the user. If there are illegal session messages, the target large language model will generate a reply message based on the context information to guide the user to output a session message that meets the message content rule. For example, "I don't think it's good to say these. Let's talk about something else."

[0196] Optionally, the target large language model can also generate a reply message for the itinerary details information according to the reply message generation rule to obtain the reply message of the target virtual object. The generation of the reply message can be expressed by formula (10) as follows:

[0197] R = φ(I d , T 1x , {T 2xi , {T 3xi}), x ∈ {n, a, m}, i ∈ {1, 2, …, K} (10)

[0198] where R represents the generated reply message, I d represents the session instruction information, T 1x , {T 2xi , {T 3xi} represents the itinerary details information, {n, a, m} respectively represent the normal scenario, the abnormal scenario, and the memory scenario, and K is the number of specific itinerary details information. Among them, the session instruction information can include the message content rule.

[0199] In an optional implementation manner, memory extraction instruction information can be obtained; the memory extraction instruction information is used to indicate the memory extraction rules that need to be satisfied during the process of the target large language model extracting memory information from the conversation messages between the target virtual object and the target role object within a preset historical time period; the target large language model performs memory extraction processing on the conversation messages within the preset historical time period according to the memory extraction rules indicated by the memory extraction instruction information to generate historical memory information.

[0200] In this embodiment, since there is a large amount of information redundancy in the conversation messages between the target role object and the target virtual object, the target large language model can be used to refine the information of different conversations. Optionally, if the current system time meets the preset system time, the memory extraction instruction information is obtained. Wherein, the current system time meeting the preset system time can mean that the current system time reaches the preset system time or the difference between the current system time and the system time of the last memory extraction is the preset system time, then memory extraction will be performed. For example, the preset system time can be zero o'clock every day, that is, when reaching zero o'clock every day, the target large language model performs memory extraction processing on the conversation messages within the preset historical time period according to the memory extraction rules indicated by the memory extraction instruction information to generate historical memory information; or, the preset system time is two hours, and every two hours, the target large language model performs memory extraction processing on the conversation messages within the preset historical time period according to the memory extraction rules indicated by the memory extraction instruction information to generate historical memory information.

[0201] Optionally, the target large language model can also perform memory extraction processing on the conversation messages within a preset historical time period according to the memory extraction rules indicated by the memory extraction instruction information and the object identifier of the target role object, to generate historical memory information. For example, the object identifier is the object name, and the information related to the target role object in the conversation messages within the preset historical time period is determined through the object name, and then memory extraction is performed on the information related to the target role object.

[0202] Optionally, the target large language model can also perform memory extraction processing on the conversation messages within a preset historical time period according to the memory extraction rules indicated by the memory extraction instruction information, the object identifier of the target role object, and the user scenario type, to generate historical memory information. For example, the object identifier is the object name, and the information related to the target role object in the conversation messages within the preset historical time period is determined through the object name. If the user scenario type is determined to be a travel scenario based on the user scenario type, then during memory extraction, accurate memory extraction will be performed according to the determined information related to the target role object and the travel scenario, and the extracted historical memory information will be better classified. For example, for the conversation message "Xingbao likes traveling the most. Do you like it, Xiaopa?", the object name is Xingbao, so the information related to Xingbao is that Xingbao likes traveling the most, and the user scenario type is travel.

[0203] Optionally, the historical memory information includes permanent historical memory information and non-permanent historical memory information; the information category of any historical memory information is determined according to the degree of association between any historical memory information and the target role object; the permanent historical memory information is permanently stored, and the non-permanent historical memory information is deleted when the corresponding expiration time is reached.

[0204] In this embodiment, the extracted historical memory information can be classified into permanent historical memory information and non-permanent historical memory information. The information category of any historical memory information is determined according to the degree of association between any historical memory information and the role object. When the degree of association with the role object is greater than or equal to the preset degree of association, it can be permanent historical memory information. For example, if the hobby of the target role object is playing basketball, the degree of association between "hobby is playing basketball" and the role object is greater than or equal to the preset degree of association, then it is permanent historical memory information. Among them, the permanent historical memory information will be permanently stored. The non-permanent historical memory information can be further classified according to the importance of the information. For example, the non-permanent historical memory information with an importance greater than or equal to the preset importance can be determined as long-term historical memory information, and the non-permanent historical memory information with an importance less than the preset importance can be determined as short-term historical memory information. Among them, the time stamp of the long-term historical memory is greater than that of the short-term historical memory. When any non-permanent historical memory information meets the expiration time indicated by the corresponding time stamp, the non-permanent historical memory information will be deleted. For example, if the conversation message of the target role object is "My hobby is playing basketball", then the permanent historical memory information can be "Hobby: playing basketball"; for example, if the conversation message of the target role object is "I want to travel recently", then the long-term historical memory information can be "Want to travel"; for example, if the conversation message of the target role object is "I want to eat Western food", then the short-term historical memory information can be "Want to eat Western food".

[0205] For example, assume that the conversation message within the preset historical time period is where d i is a round of conversation of the user or the target virtual object during the entire conversation process, the memory content is M, the corresponding memory information category is C, and the memory extraction instruction information is I M , then the memory extraction process can be expressed by formula (11) as:

[0206] M, C = φ(I m , D) (11)

[0207] Among them, C is any one of the permanent historical memory information, long-term memory information, and short-term memory information.

[0208] Optionally, if the number of session messages in the current time period is greater than or equal to a preset number of messages, obtain memory extraction instruction information; the memory extraction instruction information is used to indicate the memory extraction rules that need to be satisfied during the process of the target large language model extracting memory information based on the session messages between the target virtual object and the target role object within a preset historical time period; perform memory extraction processing on the session messages within the preset historical time period according to the memory extraction rules indicated by the memory extraction instruction information through the target large language model to generate historical memory information. Among them, the session messages within the preset historical time period include the session messages in the current time period. For example, the preset number of messages is 10. When the number of session messages between the user and the target virtual object is greater than or equal to 10, the target large language model will perform memory extraction on the session messages within the preset historical time period to obtain historical memory information.

[0209] In this embodiment, by extracting the session messages within the historical time period to obtain historical memory information and classifying the historical memory information according to the degree of association with the role object, the system can select and utilize the most relevant information, and timely discard invalid information to avoid interfering with the overall result, thereby providing a richer and more coherent user experience.

[0210] In an optional implementation manner, historical memory information and greeting instruction information can also be obtained; the greeting instruction information is used to indicate the greeting generation rules that need to be satisfied during the process of the target large language model generating greeting information based on the first object information, the second object information, the historical memory information, and at least one itinerary description information, and the historical memory information is obtained based on the historical session messages between the target virtual object and the target role object; perform greeting generation processing on the first object information, the second object information, the historical memory information, and at least one itinerary description information according to the greeting generation rules through the target large language model to generate greeting information of the target virtual object; send the greeting information to the client, and the client is used to send the greeting information in the identity of the target virtual object. Among them, the first object information may include the identity information (such as personality, type) of the target virtual object, dressing information, etc., the second object information may include the user's preference information, identity information, etc., and the historical memory information may include the user's hobbies, recent itinerary plans, etc. During the process of generating greeting information, by using the first object information, the second object information, and the historical memory information, greeting information that fits the user can be generated, and greeting information is generated in combination with at least one itinerary description information, so that the greeting information is also personalized for the target virtual object, attracting the user to interact with the target virtual object.

[0211] S410. The server sends the reply message to the client.

[0212] S411. The client sends the reply message in the identity of the target virtual object.

[0213] In the embodiments of the present application, the server performs itinerary description generation processing on the first object information, the second object information, and the itinerary outline according to the itinerary description generation rules indicated by the itinerary description instruction information through the target large language model, generates at least one itinerary description information corresponding to the itinerary outline, the server sends the at least one itinerary description information to the client, the client displays the at least one itinerary description information, and attracts the user to interact with the target virtual object through the itinerary description information, thereby enhancing user viscosity. The client sends the conversation message to the server, and the server performs reply message generation processing on the first object information, the second object information, the context information, the historical memory information, and the itinerary outline according to the reply message generation rules indicated by the conversation instruction information through the target large language model, generates the reply message of the target virtual object, and sends the reply message to the client. The client sends the reply message in the identity of the target virtual object, and further improves the user experience through the conversation between the user and the target virtual object, realizing emotional companionship for the user.

[0214] Based on the above description, please refer to Figure 5 , Figure 5 which is the management architecture diagram of the virtual object provided by the embodiments of the present application. The implementation of the interaction method of the virtual object requires access to a generative AI model or a large language model, so as to technically realize the generation of itinerary description information, real-time interaction and information intercommunication.

[0215] First, the server needs to call the target large language model to generate the robot world (BOT world) of the virtual object. The robot world (BOT world) is generated by inputting world rules, locations, interpersonal relationships, hobbies, and goals to the target large language model. Among them, the interpersonal relationships are generated by reserve NPCs.

[0216] Secondly, the server calls the target large language model to generate the itinerary outline (today's plan) of the virtual object. The today's plan is generated by inputting the open story reserve, the BOT world, and the information in the memory system to the target large language model. Among them, the open story reserve can be generated by combining the information in the memory system and the target large language model. The call of the memory system can be semi-automatic (called once a day) or fully automatic (called once every 10 rounds of conversations). The server can also record the world log (world log record) in real time, including hourly and essence-level log records, for reference and adjustment of the virtual object's behavior in real-time interaction. In addition, the server can also call the target large language model to generate the itinerary description information (itinerary expansion) of the virtual object, and generate the itinerary expansion by inputting the today's plan and the world log record to the target large language model.

[0217] The server can also manage the memory system. For example, after multiple rounds of dialogue interactions, the server can call the target large language model to generate a dialogue summary, summarizing the dialogue content so that the virtual object can remember and refer to it in subsequent conversations. For example, after every 10 rounds of dialogue, the memory system is automatically called to ensure that the virtual object can dynamically adjust its behavior and dialogue strategy during the conversation. The server can also classify memories. For example, the dialogue summaries can be classified according to their relevance to the user, which can be divided into permanent and non-permanent categories, ensuring that the memory system is organized and easy to retrieve, and regularly deleting non-permanent dialogue summaries to keep the information in the memory system concise. Optionally, the memory system can summarize on a daily and weekly basis, perform a daily log summary of today's plan to form the schedule mode of the virtual object, and perform a weekly log summary of today's plan to summarize the key events and trends of the week.

[0218] Optionally, the server can perform high-point extraction and story development. For example, at important nodes and when the story plot unfolds, the server can ensure that the virtual object can adjust the dialogue content and behavioral responses in real time based on the existing memories and world logs, enhancing the player's understanding of the virtual object's life and improving the fun of the game and the coherence of the story.

[0219] Optionally, the server can also set an opening statement. For example, at the beginning of each dialogue, the virtual object can generate an appropriate opening statement based on previous memories and the current situation to ensure the continuity and coherence of the dialogue.

[0220] Based on the above description, please refer to Figure 6 , Figure 6 which is an architecture diagram for generating itinerary description information provided by an embodiment of the present application. The server needs to access a single dialogue Agent or a large language model to technically achieve the generation, real-time interaction, and information intercommunication of itinerary description information.

[0221] The server can set up the BOT world. This includes world rules. For example, the server can define the rules and preset conditions of the virtual world to provide an infrastructure for generating itinerary description information. The server can also set up locations. For example, create locations in the virtual world. The server can also set up the interpersonal relationships, hobbies, and goals of the virtual object. For example, set up the interpersonal relationships, hobbies, and goals of the virtual object, and these factors will affect the decisions and behaviors of the virtual object in the virtual world. Optionally, the server can also perform open story reserves. For example, build an open story library for dynamically adjusting and generating new storylines according to the player's behavior. Among them, the server can also perform BOT settings. For example, set the tone - values of the virtual object, the background - world view, the personality of the virtual object (pet personality), the nickname of the virtual object (pet nickname), and the appearance setting, which can make the generated virtual object more personalized.

[0222] Optionally, the server can also set the world system according to the BOT settings and the BOT world. The server can generate NPCs based on the virtual object setting library. Through the random combination of NPCs and virtual objects, the social attributes of virtual objects can be enriched. The server can think and plan the itinerary outline and itinerary description information of virtual objects according to the BOT settings and the BOT world. The server can also unfold the story according to the BOT world and the BOT settings to attract player users to interact with virtual objects. The server can also perform high-point extraction to make the itinerary description information displayed on the client more concise and effective. The server can also generate an opening statement to ensure the continuity and coherence of the conversation. The server can also conduct target reflection to make the itinerary description information of virtual objects more in line with user interests. The server can also perform log summary to make the log generated next time more attractive for users to interact.

[0223] Optionally, after setting the virtual object, the server can perform itinerary planning, such as Daily Planning. For example, the server can generate a daily itinerary outline based on the habits, health status of the virtual object, and the goals set by the player. Specifically, the server can input the basic information of the virtual object, such as species, relationship network, and personal preferences. Then, apply the memory system to analyze the historical behaviors and reactions of the virtual object to predict its possible preferences and needs. Further, combined with the BOT settings and world rules, and considering the rules and environmental factors of the virtual world (such as weather, season, etc.), determine the daily goals and tasks that conform to the virtual object and the story development to generate a non-displayed itinerary outline as the basis for today's actions.

[0224] After generating the itinerary outline, the server can perform Itinerary Expansion, such as refining the daily plan into a specific hourly itinerary and recording the activities of the virtual object in the virtual world. Specifically, the server can extract the main activities and tasks from the daily plan, such as feeding, walking, playing, resting, etc. According to the time system of the virtual world, allocate the activities to specific hourly time periods. Generate a secondary log, that is, a secondary title, in the format of "Time: Itinerary Description" to record the player's expected actions and events.

[0225] After the journey is expanded, the server can carry out story development, such as providing a detailed description of the activities of virtual objects, enhancing players' understanding of the lives of virtual objects, and improving the fun of the game and the coherence of the story. Specifically, the server can further refine the details of each activity of the virtual object according to the hourly activities in the journey expansion. Then, natural language processing technology is applied to generate a narrative description during the activities of the virtual object, including environmental interactions, the reactions of the virtual object, etc. Optionally, the server can also record the key behaviors and growth milestones of the virtual object to form a detailed log at the minute level, that is, status information.

[0226] Optionally, the server can generate journey description information through an automated call mechanism, and the automated call mechanism can include a semi-automatic mode or a full-automatic mode. The semi-automatic mode means that the deed Log is called to generate journey description information under specific conditions (such as every day). The full-automatic mode means that the deed Log is automatically called to generate journey description information regularly (such as after every 10 rounds of conversations).

[0227] Optionally, the server can also set up a dialogue system based on BOT settings, the BOT world, user portraits, and memory retrieval. Among them, the dialogue system can achieve daily conversations, quick responses, proactive outreach, and magic constellations by calling the target large language model. Among them, the user portrait can be determined based on the user's gender, age, location, interests, social category preferences, topic tag preferences, and interactive user portrait updates. Among them, interactive user portrait updates refer to updates through memory storage, that is, the historical interactions between the user and the virtual object will update the interactive user portrait information, so that the representation of the user portrait is data-driven and more accurate. Among them, in memory retrieval, the multi-channel memory bank retrieval and screening technology is the key to achieving accurate and efficient information retrieval. The multi-channel memory bank retrieval and screening technology can include semantic similarity ranking, word frequency similarity ranking, importance ranking, time sequence ranking, and scene adaptive retrieval. Semantic similarity ranking is to use natural language processing technologies such as (term frequency-inverse document frequency, TF-IDF), Word2Vec, or (Bidirectional Encoder Representations from Transformers, BERT) to calculate the semantic similarity between the query and the entries in the memory bank, and screen the information that is semantically most relevant to the current conversation or query. TF-IDF is a commonly used weighting technology for information retrieval and data mining. Word2vec is a group of related models used to generate word vectors. BERT is a language representation model, and BERT stands for Bidirectional Encoder Representations from Transformers. Word frequency similarity ranking is to calculate the similarity between them by comparing the occurrence frequencies of texts or words in different texts and using similarity measurement methods (such as cosine similarity, Jaccard similarity, etc.), which can provide more accurate text retrieval results. Importance ranking is to assign importance scores to each piece of information in the memory bank, and then sort them based on the importance scores. Based on the type of information, user feedback, or interaction frequency, the information with higher scores is retrieved first to provide more valuable content. Time sequence ranking is to sort according to the timestamp of the information to ensure that the most recent events or interactions are considered first, achieve time sensitivity, and ensure the relevance and timeliness of the conversation. Scene adaptive retrieval is to optimize the information retrieval effect according to the user's specific scene, needs, and environmental factors. Optionally, the multi-channel memory bank retrieval and screening technology can also include memory dimension screening, that is, screening according to the memory dimensions defined in the user portrait and the virtual object portrait (such as emotions, food preferences, entertainment activities, etc.), allowing the system to select the most appropriate memory dimension according to the current context.

[0228] Optionally, the server can also memorize and store multi-round conversation interactions and world event records. Multi-round conversation interactions are obtained through the conversation system, and world event records are obtained through the world system. The memory storage system stores multiple memory nodes. For example, nodes can be obtained by summarizing multi-round conversation interactions. The server can also classify memory nodes, for example, by event or conversation; by semantic classification; and by importance level, such as importance / usefulness. The server can extract features from memory nodes to obtain semantic representations of the memory nodes, such as embeddings. The server can also summarize the conversation content and world event records of multi-round conversation interactions to obtain memory summaries. The server can also set different timestamps for memory nodes with different degrees of association based on the association between the memory node and the user. When a memory node meets the time corresponding to the timestamp, the memory node will be deleted, immediately deleting invalid data and making the stored data more concise. The timestamp can also indicate the creation and / or recent access of the memory node.

[0229] Optionally, the virtual object interaction process supports real-time conversations. For example, AI conversation generation uses advanced natural language processing technology to generate personalized conversation content based on the virtual object's behavior, emotions, and social network. Another example is World Log integration, which leverages the primary, secondary, and tertiary content in the World Log—namely, primary, secondary, and tertiary headings—to provide context for the virtual object's daily activities and special events, making conversations more relevant to pets' lives. Another example is a memory system, which combines the virtual object's historical memory with the user's current state to generate relevant and coherent conversations, enhancing the personalization and depth of the conversation. Another example is emotion recognition and expression, which analyzes input text to identify the virtual object's emotional state and expresses the corresponding emotion in the conversation, enhancing the realism of the interaction. Another example is context management, which maintains the conversation context and ensures that the virtual object's responses are consistent with previous exchanges. Another example is personalized greeting generation, where the virtual object generates personalized greetings based on the World Log and memory information to engage the user in conversation.

[0230] Optionally, the server can iterate on the dialogue effect, that is, continuously iterate and optimize the AI dialogue system through continuous dialogue effect testing and user feedback.

[0231] Optionally, the server can also perform Emoji matching, that is, using an algorithm model to match text and corresponding emotional expressions to make the conversation more vivid and expressive.

[0232] Among them, personalized greeting generation is a key function to realize the interaction between users and virtual objects. It can customize greetings according to the characteristics of virtual objects, users' behaviors, and the current context, thereby enhancing the user experience and engagement. It can include data-driven greetings, context awareness, emotional resonance, dynamic generation, diverse expressions, user behavior triggering, as well as memory and learning. Data-driven greetings mean that the generation of greetings is based on the attributes of virtual objects (such as personality, hobbies, emotional state) and the user's interaction history (such as user preferences, previous conversation content). Context awareness means that greetings should consider the current context, including time (such as morning, evening), weather, the user's location, and the current activity or state of the virtual object. Emotional resonance means that greetings should be able to express emotions and establish an emotional connection with users, such as expressing concern, joy, or comfort. Dynamic generation means that greetings are not static but are generated dynamically based on real-time data to ensure that each interaction between the user and the virtual object is novel. Diverse expressions mean that greetings should have multiple ways of expression to avoid repetition and increase the interest of the conversation. User behavior triggering means that certain user behaviors (such as long-term non-interaction, specific festivals, user birthdays, etc.) can trigger specific greetings. Memory and learning mean that the system should have a memory function to remember the key interactions between users and virtual objects and refer to these memories in future greetings.

[0233] Optionally, the server can realize the information interconnection between the user world and the virtual world. When constructing the memory system of the virtual object AI dialogue game, it is necessary to comprehensively consider different levels of logging and context information, as well as how to effectively integrate and retrieve this information, specifically including context understanding, memory bank construction, and KV pair storage. Context understanding can include memory system context and current conversation context. The memory system context can include hourly and minute-level logs, recording the detailed timeline of the daily activities of the pet and user interactions. Minute-level logs provide detailed event records, such as the behaviors of virtual objects or user instructions at specific time points. Hourly logs summarize the key activities and interactions over a period of time for quick review. The current conversation context refers to real-time tracking of the ongoing conversation and understanding the flow and theme of the conversation to generate coherent and relevant responses. Real-time tracking of the ongoing conversation includes the communication content between the user and the virtual object.

[0234] Optionally, the server can also construct a memory bank. The memory bank includes daily / weekly logs. The logs store the behavior and interaction patterns of virtual objects and users over a long time span. Daily logs record the main events and activities of each day, forming the daily pattern of virtual objects. Weekly logs provide a more macroscopic perspective, summarizing the key events and trends within a week. The server can also generate a conversation summary, that is, after the conversation ends, extract the key information and summary and store them in the memory bank for future reference.

[0235] Optionally, the server can also store the user profile and the virtual object profile in key-value pairs respectively. For example, store the user's basic information, preferences, behavior patterns, etc. in the form of key-value pairs (Key-Value Pairs). Exemplarily, {"User ID": "U123", "Hobbies": ["Music", "Reading"], "Recent Mood": "Happy"}. Another example is to store information such as the pet's personality traits, preferences, and social relationships. Exemplarily, {"Pet ID": "P456", "Personality": "Lively", "Favorite Game": "Hide and Seek"}.

[0236] The embodiment of the present application also provides a computer storage medium, in which program instructions are stored. When the program instructions are executed, they are used to implement the corresponding methods described in the above embodiments.

[0237] The embodiment of the present application provides a computer program product. The computer program product includes a computer program, and the computer program is stored in a computer storage medium; the processor of the computer device reads the computer program from the computer storage medium, and the processor executes the computer program, so that the computer device executes the corresponding methods described in the above embodiments.

[0238] See again Figure 7 , Figure 7 which is a schematic structural diagram of an interactive device for a virtual object provided by an embodiment of the present application.

[0239] In one implementation of the interactive device for a virtual object according to the embodiment of the present application, the interactive device for a virtual object includes the following structure:

[0240] An acquisition unit 701, configured to acquire first object information of a target virtual object, second object information of a target role object, and outline instruction information; the outline instruction information is used to indicate an outline generation rule;

[0241] A generation unit 702, configured to perform outline generation processing on the first object information and the second object information according to the outline generation rule through a target large language model, and generate an itinerary outline of the target virtual object within a preset time period;

[0242] The acquisition unit 701 is further configured to acquire itinerary description instruction information, and the itinerary description instruction information is used to indicate an itinerary description generation rule;

[0243] The generation unit 702 is further configured to perform itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rule through a target large language model, and generate at least one itinerary description information corresponding to the itinerary outline;

[0244] A sending unit 703 is configured to send at least one travel description message to a client corresponding to a target role object, and the client is configured to display the at least one travel description message.

[0245] In one implementation, a generating unit 702 performs travel outline generation processing on first object information and second object information according to an outline generation rule through a target large language model, and generates a travel outline of a target virtual object within a preset time period, which can be used for:

[0246] Performing travel outline processing on at least one of the first object information, the second object information, historical memory information, and the social relationship between the target virtual object and other virtual objects according to the outline generation rule indicated by the outline instruction information through the target large language model, and generating a travel outline of the target virtual object within a preset time period; the historical memory information is obtained based on historical conversation messages between the target virtual object and the target role object.

[0247] In one implementation, an obtaining unit 701 can also be used for:

[0248] Obtaining travel details instruction information, where the travel details instruction information is used to indicate a travel details generation rule;

[0249] Performing travel details generation processing on the first object information, historical memory information, travel outline, and any travel description message according to the travel details generation rule through the target large language model, and generating travel details information corresponding to any travel description message.

[0250] In one implementation, a generating unit 702 performs travel details generation processing on the first object information, historical memory information, travel outline, and any travel description message according to the travel details generation rule through the target large language model, and generates travel details information corresponding to any travel description message, which can be used for:

[0251] Obtaining an initial travel time interval corresponding to any travel description message;

[0252] Based on a preset time constraint, performing a shrinking process on the initial travel time interval to obtain a target travel time interval;

[0253] Performing travel details generation processing on the first object information, historical memory information, travel outline, target travel time interval, and any travel description message according to the travel details generation rule through the target large language model, and generating travel details information corresponding to any travel description message; the travel content corresponding to the travel details information is within the target travel time interval.

[0254] In one implementation, the travel details instruction information includes a travel details example corresponding to the travel description message, and the travel details example includes a target corresponding to the travel description message.

[0255] In one implementation, the generation unit 702 performs a trip description generation process on the first object information and the trip outline according to the trip description generation rules through the target large language model, and generates at least one trip description information corresponding to the trip outline, which can be used for:

[0256] Obtain the attribute information of the target virtual object, where the attribute information is used to indicate the object behavior of the target virtual object; perform a trip description generation process on the first object information, the attribute information, and the trip outline according to the trip description generation rules through the target large language model, and generate at least one trip description information corresponding to the trip outline; or

[0257] Obtain the consensus information and the content order of the trip content corresponding to each trip description information in the at least one trip description information; perform a trip description generation process on the first object information, the consensus information, the content order, and the trip outline according to the trip description generation rules through the target large language model, and generate at least one trip description information corresponding to the trip outline.

[0258] In one implementation, the acquisition unit 701 can also be used for:

[0259] Obtain the session instruction information, the historical memory information, and the context information. The session instruction information is used to indicate the reply message generation rules, and the context information is used to indicate the session messages between the target virtual object and the target role object before the sending time of the target session message, and the session messages sent by the target role object after the sending time of the target session message. The time difference between the sending time of the target session message and the current system time is less than or equal to the sending time of other session messages sent by the target virtual object, and the target session message is sent in the identity of the target virtual object;

[0260] Perform a reply message generation process on the first object information, the second object information, the context information, the historical memory information, and the trip outline according to the reply message generation rules through the target large language model, and generate a reply message of the target virtual object;

[0261] Send the reply message to the client, and the client is used to display the reply message in the identity of the target virtual object.

[0262] In one implementation, the acquisition unit 701 can also be used for:

[0263] Obtain the memory extraction instruction information; the memory extraction instruction information is used to indicate the memory extraction rules required for the target large language model to extract memory information according to the session messages between the target virtual object and the target role object in a preset historical time period.

[0264] The session messages within a preset historical time period are processed for memory extraction by the target large language model according to the memory extraction rules to generate historical memory information.

[0265] In one implementation, the historical memory information includes permanent historical memory information and non-permanent historical memory information; the information category of any historical memory information is determined according to the degree of association between any historical memory information and the target role object; the permanent historical memory information is permanently stored, and the non-permanent historical memory information is deleted when the corresponding expiration time is reached.

[0266] In one implementation, the acquisition unit 701 can also be used for:

[0267] Acquire the fourth object information of the training virtual object, the fifth object information of the training role object, the preset itinerary outline, and the outline instruction information; the outline instruction information is used to indicate the outline generation rules;

[0268] The fourth object information and the fifth object information are processed for itinerary outline generation by the initial large language model according to the outline generation rules to generate the itinerary outline of the training virtual object within the preset time period;

[0269] With the goal of reducing the difference between the itinerary outline of the training virtual object within the preset time period and the preset itinerary outline, the initial large language model is trained, and after the training is completed, the target large language model is obtained; the target large language model is used to process the first object information and the second object information for itinerary outline generation according to the outline generation rules to generate the itinerary outline of the target virtual object within the preset time period.

[0270] In the embodiments of the present application, an acquisition unit 701 acquires first object information of a target virtual object, second object information of a target role object, and outline instruction information, where the outline instruction information is used to indicate an outline generation rule; a generation unit 702 performs itinerary outline generation processing on the first object information and the second object information according to the outline generation rule through a target large language model to generate an itinerary outline of the target virtual object within a preset time period; the acquisition unit 701 acquires itinerary description instruction information, where the itinerary description instruction information is used to indicate an itinerary description generation rule; the generation unit 702 performs itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rule through the target large language model to generate at least one itinerary description information corresponding to the itinerary outline; a sending unit 703 sends the at least one itinerary description information to a client corresponding to the target role object, and the client is used to display the at least one itinerary description information. In the embodiments of the present application, the itinerary outline and the itinerary description information are generated hierarchically, the itinerary content is gradually refined, and combined with the generation capabilities of the large language model for different tasks, the itinerary description information of the target virtual object for generating text content can be generated at a relatively low cost, realizing lightweight generation of interactive content of virtual objects. In addition, even when the user does not initiate an interaction instruction, the virtual pet will actively generate corresponding interactive content, which can attract the user to interact with the virtual object to enhance user viscosity.

[0271] See again Figure 8 , Figure 8 FIG. is a schematic structural diagram of a computer device provided by an embodiment of the present application. The computer device in the embodiment of the present application includes structures such as a power supply module, and includes a processor 801, a memory 802, and a communication interface 803. Data can be exchanged between the processor 801, the memory 802, and the communication interface 803, and the processor 801 implements the interactive method of the corresponding virtual object.

[0272] The memory 802 may include a volatile memory, such as a random-access memory (RAM); the memory 802 may also include a non-volatile memory, such as a flash memory, a solid-state drive (SSD), etc.; the memory 802 may further include a combination of the above types of memories.

[0273] The processor 801 can be a central processing unit (CPU). The processor 801 can also be a combination of a CPU and a GPU. In a computer device, multiple CPUs and GPUs can be included as needed to perform interactions of corresponding virtual objects. In one embodiment, the memory 802 is used to store program instructions. The processor 801 can call the program instructions to implement various methods involved in the above embodiments of the present application.

[0274] In the first possible implementation manner, the processor 801 of the computer device calls the program instructions stored in the memory 802 to obtain the first object information of the target virtual object, the second object information of the target role object, and the outline instruction information; the outline instruction information is used to indicate the outline generation rule; the target large language model performs outline generation processing on the first object information and the second object information according to the outline generation rule to generate an itinerary outline of the target virtual object within a preset time period; obtain itinerary description instruction information, where the itinerary description instruction information is used to indicate the itinerary description generation rule; the target large language model performs itinerary description generation processing on the first object information and the itinerary outline according to the itinerary description generation rule to generate at least one itinerary description information corresponding to the itinerary outline; send the at least one itinerary description information to the client corresponding to the target role object, and the client is used to display the at least one itinerary description information.

[0275] In one embodiment, when the processor 801 performs itinerary outline generation processing on the first object information and the second object information according to the outline generation rule through the target large language model to generate an itinerary outline of the target virtual object within a preset time period, the following operations can be performed:

[0276] The target large language model performs itinerary outline processing on at least one of the first object information, the second object information, the historical memory information, and the social relationship between the target virtual object and other virtual objects according to the outline generation rule indicated by the outline instruction information to generate an itinerary outline of the target virtual object within a preset time period; the historical memory information is obtained based on the historical conversation messages between the target virtual object and the target role object.

[0277] In one embodiment, the processor 801 can also perform the following operations:

[0278] Obtain itinerary details instruction information, where the itinerary details instruction information is used to indicate the itinerary details generation rule;

[0279] The target large language model performs itinerary details generation processing on the first object information, the historical memory information, the itinerary outline, and any one of the itinerary description information according to the itinerary details generation rule to generate itinerary details information corresponding to any one of the itinerary description information.

[0280] In one implementation, the processor 801 performs trip details generation processing on the first object information, historical memory information, trip outline, and any trip description information according to the trip details generation rules of the target large language model to generate trip details information corresponding to any trip description information, and can perform the following operations:

[0281] Obtain the initial trip time interval corresponding to any trip description information;

[0282] Based on the preset time constraints, perform a reduction process on the initial trip time interval to obtain the target trip time interval;

[0283] Perform trip details generation processing on the first object information, historical memory information, trip outline, target trip time interval, and any trip description information according to the trip details generation rules of the target large language model to generate trip details information corresponding to any trip description information; the trip content corresponding to the trip details information is carried out within the target trip time interval.

[0284] In one implementation, the trip details instruction information includes a trip details example corresponding to the trip description information, and the trip details example includes the target corresponding to the trip description information.

[0285] In one implementation, the processor 801 performs trip description generation processing on the first object information and the trip outline according to the trip description generation rules of the target large language model to generate at least one trip description information corresponding to the trip outline, and can perform the following operations:

[0286] Obtain the attribute information of the target virtual object, where the attribute information is used to indicate the object behavior of the target virtual object; perform trip description generation processing on the first object information, attribute information, and trip outline according to the trip description generation rules of the target large language model to generate at least one trip description information corresponding to the trip outline; or

[0287] Obtain the consensus information and content order of the trip content corresponding to each trip description information in at least one trip description information; perform trip description generation processing on the first object information, consensus information, content order, and trip outline according to the trip description generation rules of the target large language model to generate at least one trip description information corresponding to the trip outline.

[0288] In one implementation, the processor 801 can also perform the following operations:

[0289] Acquire conversation instruction information, historical memory information, and context information, where the conversation instruction information indicates a reply message generation rule, and the context information indicates conversation messages between the target virtual object and the target character object before the sending time of the target conversation message, as well as conversation messages sent by the target character object after the sending time of the target conversation message. The time difference between the sending time of the target conversation message and the current system time is less than or equal to the sending time of other conversation messages sent by the target virtual object, and the target conversation message is sent as the target virtual object.

[0290] Performing reply message generation processing on the first object information, the second object information, the context information, the historical memory information, and the itinerary outline using the target large language model according to the reply message generation rule to generate a reply message for the target virtual object;

[0291] The reply message is sent to the client, and the client is used to display the reply message as the target virtual object.

[0292] In one implementation, the processor 801 may further perform the following operations:

[0293] Obtaining memory extraction instruction information; the memory extraction instruction information is used to instruct the target large language model to satisfy the memory extraction rules that need to be satisfied in the process of extracting memory information based on the conversation messages between the target virtual object and the target character object within a preset historical time period;

[0294] The target large language model performs memory extraction processing on conversation messages within a preset historical time period according to the memory extraction rules to generate historical memory information.

[0295] In one embodiment, historical memory information includes permanent historical memory information and non-permanent historical memory information; the information category of any historical memory information is determined based on the association between any historical memory information and the target role object; permanent historical memory information is permanently stored, and non-permanent historical memory information is deleted when the corresponding expiration time is reached.

[0296] In one implementation, the processor 801 may further perform the following operations:

[0297] Acquire fourth object information of the training virtual object, fifth object information of the training character object, a preset itinerary outline, and outline instruction information; the outline instruction information is used to indicate outline generation rules;

[0298] Performing itinerary outline generation processing on the fourth object information and the fifth object information according to outline generation rules using the initial large language model to generate a itinerary outline for the training virtual object within a preset time period;

[0299] The initial large language model is trained with the goal of reducing the difference between the travel outline of the training virtual object within a preset time period and the preset travel outline, and the target large language model is obtained after the training is completed; the target large language model is used to perform travel outline generation processing on the first object information and the second object information according to the outline generation rule, and generate the travel outline of the target virtual object within the preset time period.

[0300] In the embodiment of the present application, the processor 801 obtains the first object information of the target virtual object, the second object information of the target role object, and the outline instruction information, where the outline instruction information is used to indicate the outline generation rule; the target large language model performs travel outline generation processing on the first object information and the second object information according to the outline generation rule, and generates the travel outline of the target virtual object within the preset time period; obtains the travel description instruction information, where the travel description instruction information is used to indicate the travel description generation rule; the target large language model performs travel description generation processing on the first object information and the travel outline according to the travel description generation rule, and generates at least one travel description information corresponding to the travel outline; sends the at least one travel description information to the client corresponding to the target role object, and the client is used to display the at least one travel description information. In the embodiment of the present application, the travel outline and the travel description information are generated hierarchically, the travel content is gradually refined, and the generation ability of the large language model for different tasks is combined, so that the travel description information of the target virtual object with text content can be generated at a lower cost, and the lightweight generation of the interactive content of the virtual object is realized. In addition, even when the user does not initiate an interaction instruction, the virtual pet will actively generate corresponding interactive content, which can attract the user to interact with the virtual object to enhance user viscosity.

[0301] Those of ordinary skill in the art can understand all or part of the processes in the above method embodiments. This process can be completed by relevant hardware instructed by a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. The foregoing storage media include: various media such as ROM or random access memory RAM, magnetic disks, or optical discs that can store program codes.

[0302] The above-disclosed are only some embodiments of the present application. Of course, the scope of rights of the present application cannot be limited by this. Those of ordinary skill in the art can understand all or part of the processes of the above embodiments, and the equivalent changes made according to the claims of the present application still fall within the scope covered by the present application.

Claims

1. An interactive method for virtual objects, characterized in that, The method includes: Obtaining first object information of a target virtual object, second object information of a target role object, and outline instruction information; the outline instruction information is used to indicate an outline generation rule; Performing itinerary outline generation processing on the first object information and the second object information by the target large language model according to the outline generation rule, to generate an itinerary outline of the target virtual object within a preset time period; Obtaining itinerary description instruction information, where the itinerary description instruction information is used to indicate an itinerary description generation rule; Performing itinerary description generation processing on the first object information and the itinerary outline by the target large language model according to the itinerary description generation rule, to generate at least one itinerary description information corresponding to the itinerary outline; Sending the at least one itinerary description information to a client corresponding to the target role object, where the client is used to display the at least one itinerary description information.

2. The method according to claim 1, wherein The performing itinerary outline generation processing on the first object information and the second object information by the target large language model according to the outline generation rule, to generate an itinerary outline of the target virtual object within a preset time period, includes: Performing itinerary outline processing on at least one of the first object information, the second object information, historical memory information, and the social relationship between the target virtual object and other virtual objects by the target large language model according to the outline generation rule indicated by the outline instruction information, to generate an itinerary outline of the target virtual object within a preset time period; the historical memory information is obtained based on historical conversation messages between the target virtual object and the target role object.

3. The method according to claim 1, characterized in that, The method further includes: Obtaining itinerary details instruction information, where the itinerary details instruction information is used to indicate an itinerary details generation rule; Performing itinerary details generation processing on the first object information, the historical memory information, the itinerary outline, and any itinerary description information by the target large language model according to the itinerary details generation rule, to generate itinerary details information corresponding to the any itinerary description information.

4. The method according to claim 3, wherein The performing itinerary details generation processing on the first object information, the historical memory information, the itinerary outline, and any itinerary description information by the target large language model according to the itinerary details generation rule, to generate itinerary details information corresponding to the any itinerary description information, includes: Obtaining an initial itinerary time interval corresponding to the any itinerary description information; Performing a shrinking process on the initial itinerary time interval based on a preset time constraint, to obtain a target itinerary time interval; Performing itinerary details generation processing on the first object information, the historical memory information, the itinerary outline, the target itinerary time interval, and the any itinerary description information by the target large language model according to the itinerary details generation rule, to generate itinerary details information corresponding to the any itinerary description information; the itinerary content corresponding to the itinerary details information is within the target itinerary time interval.

5. The method according to claim 4, characterized in that The itinerary details instruction information includes an itinerary details example corresponding to the itinerary description information, and the itinerary details example includes a target corresponding to the itinerary description information.

6. The method according to claim 1, wherein The process of generating at least one itinerary description information corresponding to the itinerary outline by the target large language model according to the itinerary description generation rule for the first object information and the itinerary outline includes: Obtaining the attribute information of the target virtual object, where the attribute information is used to indicate the object behavior of the target virtual object; generating at least one itinerary description information corresponding to the itinerary outline by the target large language model according to the itinerary description generation rule for the first object information, the attribute information, and the itinerary outline; or Obtaining the consensus information and content order of the itinerary content corresponding to each itinerary description information in the at least one itinerary description information; generating at least one itinerary description information corresponding to the itinerary outline by the target large language model according to the itinerary description generation rule for the first object information, the consensus information, the content order, and the itinerary outline.

7. The method according to claim 1, wherein The method further includes: Obtaining session instruction information, historical memory information, and context information, where the session instruction information is used to indicate a reply message generation rule, and the context information is used to indicate the session messages between the target virtual object and the target role object before the sending time of the target session message, and the session messages sent by the target role object after the sending time of the target session message. The time difference between the sending time of the target session message and the current system time is less than or equal to the sending time of other session messages sent by the target virtual object, and the target session message is sent in the identity of the target virtual object; Generating a reply message of the target virtual object by the target large language model according to the reply message generation rule for the first object information, the second object information, the context information, the historical memory information, and the itinerary outline; Sending the reply message to the client, and the client is used to display the reply message in the identity of the target virtual object.

8. The method according to claim 1, wherein The method further includes: Obtaining memory extraction instruction information; the memory extraction instruction information is used to indicate the memory extraction rules required to be satisfied by the target large language model when extracting memory information according to the session messages between the target virtual object and the target role object in a preset historical period; Performing memory extraction processing on the session messages in the preset historical period by the target large language model according to the memory extraction rules to generate historical memory information.

9. The method according to claim 8, wherein The historical memory information includes permanent historical memory information and non-permanent historical memory information; the information category of any historical memory information is determined according to the association degree between the any historical memory information and the target role object; the permanent historical memory information is permanently stored, and the non-permanent historical memory information is deleted when the corresponding expiration time is reached.

10. The method according to claim 1, wherein The method further includes: Obtaining fourth object information of a training virtual object, fifth object information of a training character object, a preset itinerary outline, and outline instruction information; the outline instruction information is used to indicate an outline generation rule; Performing itinerary outline generation processing on the fourth object information and the fifth object information by the initial large language model according to the outline generation rule to generate an itinerary outline of the training virtual object within a preset time period; Training the initial large language model with the goal of reducing the difference between the itinerary outline of the training virtual object within the preset time period and the preset itinerary outline, and obtaining the target large language model after the training is completed; the target large language model is used to perform itinerary outline generation processing on the first object information and the second object information according to the outline generation rule to generate an itinerary outline of the target virtual object within the preset time period.

11. An interactive system for virtual objects, characterized in that, The system includes a client and a server, and the server includes an itinerary service module and an itinerary generation module, where: The client is used to send an itinerary acquisition request to the server; The itinerary service module is used to, in response to the itinerary acquisition request, obtain at least one itinerary description information from the itinerary generation module and send the at least one itinerary description information to the client; The client is further used to display at least one itinerary description information; The itinerary generation module is used to perform itinerary outline generation processing on the first object information and the second object information by the target large language model according to the outline generation rule to generate an itinerary outline of the target virtual object within a preset time period, and perform itinerary description generation processing on the first object information and the itinerary outline by the target large language model according to the itinerary description generation rule to generate at least one itinerary description information corresponding to the itinerary outline.

12. The system according to claim 11, wherein The server further includes a dialogue service module and / or a memory summary module, where: The dialogue service module is used to, in response to a dialogue reply request, obtain historical memory information and an itinerary outline, and perform reply message generation processing on the first object information, the second object information, context information, the historical memory information, and the itinerary outline by the target large language model according to the reply message generation rule to generate a reply message of the target virtual object; The memory summary module is used to perform memory extraction processing on session messages within a preset historical time period by the target large language model according to the memory extraction rule to generate historical memory information.

13. An interactive device for virtual objects, characterized in that, Including: An obtaining unit, used to obtain first object information of a target virtual object, second object information of a target character object, and outline instruction information; The outline instruction information is used to indicate an outline generation rule; A generating unit, used to perform itinerary outline generation processing on the first object information and the second object information by the target large language model according to the outline generation rule to generate an itinerary outline of the target virtual object within a preset time period; The obtaining unit is further used to obtain itinerary description instruction information, and the itinerary description instruction information is used to indicate an itinerary description generation rule; The generating unit is further configured to perform a trip description generation process on the first object information and the trip outline according to the trip description generation rule through the target large language model, and generate at least one trip description information corresponding to the trip outline; The sending unit is configured to send the at least one trip description information to the client corresponding to the target role object, and the client is configured to display the at least one trip description information.

14. A computer device, characterized in that, The computer device includes a memory, a communication interface, and a processor, wherein the memory, the communication interface, and the processor are interconnected; the memory stores a computer program, and the processor calls the computer program stored in the memory to implement the interactive method of the virtual object according to any one of claims 1 to 10.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the interactive method of the virtual object according to any one of claims 1 to 10.

16. A computer program product, characterized in that, The computer program product includes a computer program, and the computer program is stored in a computer storage medium; a processor of a computer device reads the computer program from the computer storage medium, and the processor executes the computer program, so that the computer device executes the interactive method of the virtual object according to any one of claims 1 to 10.

Images