Method and system for generating figure experience of digital person

Abstract

The invention discloses a figure experience generation method and system for a digital person, and the method comprises the steps: obtaining the basic setting information based on the digital person inputted by a user, and carrying out the analysis of the basic setting information, and obtaining key element information; querying related information matched with the key element information in a knowledge material library, and inputting the key element information and the related information into an experience generation model to obtain primary experience information; and performing secondary processing on the primary experience information to obtain secondary experience information, and formatting and outputting the secondary experience information according to a preset format. According to the method, the primary experience information of the digital person is generated through the large model, and secondary processing such as verification and expansion is performed on the primary experience information, so that the logicality and diversity figure experience which conforms to personalized figure setting of the digital person can be efficiently and intelligently generated, and the reality sense, attraction and application value of the digital person are further improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a method and system for generating the experiences of a digital human. Background Technology

[0002] As a product of the fusion of artificial intelligence and computer graphics, digital humans have developed rapidly in recent years and have been widely applied in various fields. From virtual anchors to intelligent customer service representatives, and then to virtual avatars in the metaverse, digital humans are gradually permeating people's daily lives and work. A captivating and persuasive digital human not only needs a realistic appearance and fluent communication skills, but also must possess unique and rich "personal experiences" to construct its personality, background, and behavioral patterns.

[0003] Currently, the generation of digital human character experiences mainly relies on manual writing and template-based generation. Manual writing involves professional screenwriters or content creators manually crafting character stories and background experiences based on the digital human's established characteristics (such as identity, personality, and environment). While this method ensures the quality and depth of the experiences, it is time-consuming, labor-intensive, costly, and difficult to achieve large-scale, rapid personalized customization. Pre-setting some experience templates or fixed logical rules and generating character experiences by combining different elements can improve efficiency, but the generated content is often rigid, lacks originality and realism, and fails to meet users' requirements for depth and complexity in digital human creation.

[0004] With the rise of large-scale models, their powerful text understanding, logical reasoning, and content generation capabilities offer new possibilities for solving the aforementioned problems. Large-scale models can learn from massive amounts of text data, thereby acquiring rich world knowledge and narrative patterns. Some traditional large-scale language models can generate textual descriptions of digital human experiences based on keywords or simple settings. However, these models have limitations in understanding complex contexts, maintaining long-term consistency, and generating long-form experiences with creative and emotional depth.

[0005] Existing methods merely use large models as simple text-filling tools, lacking in-depth control over the personality and coherence of digital humans' experiences. This results in generated experiences that may not match the digital human's established character, or plot developments that are abrupt and lack depth. Therefore, how to effectively utilize large models to generate character experiences for digital humans that are both consistent with their specific persona and possess logic, diversity, and appeal remains a challenging issue. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for generating digital human character experiences, aiming to efficiently and intelligently generate character experiences that conform to the personalized persona of the digital human, are logical and diverse, thereby enhancing the realism, attractiveness and application value of the digital human.

[0007] This invention discloses a method for generating the experiences of a digital human, comprising:

[0008] Obtain basic settings information based on digital humans input by users, and parse the basic settings information to obtain key element information;

[0009] Search the knowledge resource library for relevant information that matches the key elements. The relevant information includes background information, plot fragments, and narrative logic.

[0010] Input key element information and related information into the experience generation model, and obtain the initial experience information output by the experience generation model;

[0011] The primary experience information is processed to obtain secondary experience information. The secondary processing is used to verify and expand the primary experience information.

[0012] The secondary experience information is formatted according to a preset format, and the formatted secondary experience information is output.

[0013] Optionally, the basic setup information includes at least one of the following: identity tags, personality traits, key skills, desired themes, and specific event nodes.

[0014] Optionally, the basic setting information is parsed to obtain key element information, including:

[0015] The basic setting information is preprocessed, including cleaning, splitting, and standardization.

[0016] The preprocessed basic information is structured and parsed using a large language model to obtain key element information.

[0017] Optionally, the knowledge resource library includes at least one of the following: general world knowledge, domain-specific knowledge, narrative structure templates, plot elements, and emotional patterns.

[0018] Optionally, querying the knowledge resource base for relevant information matching the key element information includes:

[0019] Retrieve multidimensional information corresponding to key elements from the knowledge resource database;

[0020] By integrating multi-dimensional information, relevant information can be obtained.

[0021] Optionally, the generative model may include at least one pre-trained large language model;

[0022] The methods for generating digital human experiences also include:

[0023] The preset prompting engineering control experience generation model is used to output primary experience information, which is used to represent the digital human's character experience framework and key plots.

[0024] The prompting process includes a first prompt word, which is used to instruct the experience generation model to generate a character experience outline.

[0025] Optionally, the primary experience information is further processed to obtain secondary experience information, including:

[0026] Primary experience information and key element information are input into the secondary model, which includes at least one pre-trained large language model.

[0027] The prompting engineering control model performs secondary processing on primary experience information. The prompting engineering includes a second prompt word, a third prompt word, and a fourth prompt word. The second prompt word is used to perform logical consistency verification on the primary experience information; the third prompt word is used to perform personalized enhancement on the primary experience information; and the fourth prompt word is used to perform diversity and innovative expansion on the primary experience information.

[0028] Obtain the secondary experience information output by the secondary model.

[0029] Optionally, a secondary processing method is used to perform secondary processing on the primary experience information using a prompting engineering control secondary model, including:

[0030] Verify whether the initial experience information contradicts the basic settings information, and output the verification results, including whether the settings contradict each other and whether the settings match.

[0031] Refine the behaviors, dialogues, or decisions in the initial experience information based on the basic setting information;

[0032] Decision points based on primary experience information provide at least one alternative path to expand upon the primary experience information.

[0033] Optionally, the preset format includes at least one of story text, time series, dialogue script, and structured data.

[0034] This invention also discloses a digital human character experience generation system, used to execute the above-described digital human character experience generation method. The digital human character experience generation system includes:

[0035] The input parsing module is configured to: acquire basic setting information based on the digital human input by the user, and parse the basic setting information to obtain key element information;

[0036] The material integration module is configured to: query relevant information in the knowledge material library that matches the key element information, including background information, plot fragments, and narrative logic;

[0037] The core generation module is configured to: input key element information and related information into the experience generation model, and obtain the primary experience information output by the experience generation model;

[0038] The secondary processing module is configured to perform secondary processing on the primary experience information to obtain secondary experience information. The secondary processing is used to verify and expand the primary experience information.

[0039] The data output module is configured to format the secondary experience information according to a preset format and output the formatted secondary experience information.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] 1. High degree of automation and efficiency improvement: By leveraging the generation capabilities of large models, the workload of manually compiling character experiences is significantly reduced, greatly improving the efficiency of experience generation and enabling the rapid creation of background stories for a large number of digital humans.

[0042] 2. Deep personalization: It can generate highly personalized character experiences based on detailed settings (identity, personality, skills, etc.) input by users, so that each digital human has a unique "life" and enhances the uniqueness and realism of digital humans.

[0043] 3. Enhanced richness and creativity: The large model has mastered a massive amount of knowledge and narrative patterns, enabling it to generate richer, more diverse, and more creative experience content. This overcomes the limitations of traditional template-based generation. Furthermore, through secondary processing, it ensures that the generated experiences are logically self-consistent and consistent with the core character settings, thereby enhancing the credibility of the experiences.

[0044] 4. High flexibility and scalability: Users can flexibly control the style and direction of the generated experience by adjusting input parameters and desired themes. The modular design of the system also facilitates future integration with more advanced large models or richer knowledge bases.

[0045] 5. Lowering the barriers to creation and costs: This allows developers or users without professional screenwriting skills to create high-quality character experiences for digital humans, thus lowering the barriers to digital content creation and costs. Attached Figure Description

[0046] Figure 1 A flowchart illustrating the method for generating the digital human's life experiences provided by this invention;

[0047] Figure 2This is a schematic diagram of the structure of the digital human character experience generation system provided by the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] The present invention will now be described in further detail with reference to the accompanying drawings.

[0050] like Figure 1 As shown, this embodiment of the invention provides a method for generating the experiences of a digital human, including the following steps:

[0051] Step S1: Obtain the basic settings information based on the digital human input by the user, and parse the basic settings information to obtain key element information.

[0052] In this embodiment of the invention, the basic setting information includes at least one of identity tags, personality traits, key skills, desired experience themes, and specific event nodes. Thus, by acquiring the user-inputted basic setting information based on the digital human, rich material can be provided for generating the digital human's character experience; the digital human's identity background can be set, making it more three-dimensional and realistic; the characterization of personality traits makes the digital human's performance more consistent with the preset image; the introduction of key skills gives the digital human specific abilities, increasing its interactivity and playability in the virtual world; and the setting of desired experience themes and specific event nodes provides a clear direction and framework for generating the digital human's character experience, making the generated storyline more reasonable.

[0053] For example, identity labels can be the digital person's profession (such as doctor, artist, or detective), social role (mother, student, or volunteer), and historical context (contemporary, Republican era, or near future). Personality traits can be expressed using standardized scale values ​​(such as specific values ​​for the Big Five personality traits: openness, conscientiousness, extraversion, agreeableness, neuroticism), or they can be more descriptive natural language labels (such as: optimistic, introverted, decisive, compassionate, cynical).

[0054] Key skills could include fluency in multiple languages, proficiency in programming, exceptional driving skills, or extraordinary insight. The desired experience theme can be the user's designated main tone (e.g., inspirational growth, joy and sorrow, suspenseful exploration). Specific time points can be provided to the user, including key event milestones (e.g., experiencing a major failure, meeting a life-changing mentor, uncovering a family secret).

[0055] Furthermore, the process of parsing the basic setting information to obtain key element information is as follows: First, the basic setting information is preprocessed, and then the preprocessed basic setting information is structured and parsed using a large language model to obtain the key element information. The preprocessing includes cleaning, splitting, and standardization. The parsed key element information is structured into JSON or other formats and passed to subsequent modules.

[0056] The cleaning step removes redundant and erroneous information from the basic settings, the splitting step breaks down complex information into smaller, easier-to-process units, and the standardization step unifies the format and expression of information, making the generated key element information easier for large language models to understand.

[0057] For example, the system uses its built-in natural language processing model to perform entity recognition, keyword extraction, and sentiment analysis on the input basic information. For instance, if the input is "An introverted female forensic doctor living in 21st-century Shanghai, she is calm and observant, and we hope her story revolves around how she overcomes the confusion of starting her career and solves a difficult case," the system will parse out the following: "Core Identity": "Forensic Doctor," "Geographical Background": "Shanghai," "Time Period": "21st Century," "Personality": "Introverted, calm, observant," "Expected Theme": "Growth, Suspense," and "Key Plot Points": "Overcoming workplace confusion and solving a difficult case."

[0058] Step S2: Query the knowledge material library for relevant information that matches the key element information.

[0059] In this embodiment of the invention, the knowledge resource library provides factual basis and creative inspiration for the generation of digital human task experiences. The knowledge resource library includes at least one of the following: general world knowledge, domain-specific knowledge, narrative structure templates, plot elements, and emotional patterns.

[0060] For example, general world knowledge encompasses common sense, historical events, geography and culture, and scientific principles. For instance, for a setting like the "Republic of China era," information such as the social landscape, major historical events, and price levels of that period could be retrieved. Professional domain knowledge includes in-depth knowledge specific to a particular profession or background. For example, for a "forensic" setting, it could include forensic procedures, common case types, and professional terminology. Narrative materials include classic narrative structures (such as a hero's journey or a three-act structure), plot archetypes (such as encounters, conflicts, and turning points), emotional development curves, and character relationship templates.

[0061] In this embodiment of the invention, multi-dimensional information corresponding to key element information is retrieved from the knowledge material library, and the multi-dimensional information is integrated to obtain relevant information, including background information, plot fragments, and narrative logic.

[0062] This invention automatically triggers a search based on the extracted key elements, retrieving and integrating relevant background information, plot fragments, and narrative logic from a knowledge base. For example, based on the settings of "forensic medicine" and "Shanghai," the system will retrieve and integrate information such as "the organizational structure of existing forensic identification institutions in Shanghai," "the daily work of forensic doctors," and "the characteristics of criminal cases in Shanghai during specific periods" as "fact anchors" for the generated content, ensuring the authenticity and professionalism of the experience.

[0063] Step S3: Input the key element information and related information into the experience generation model, and obtain the primary experience information output by the experience generation model.

[0064] In this embodiment of the invention, the experience generation model includes at least one pre-trained large language model. The method for generating the character experience of a digital human includes: using a preset cueing engineering to control the experience generation model to output preliminary experience information, which is used to represent the character experience framework and key plots of the digital human; the cueing engineering includes a first cue word, which is used to instruct the experience generation model to generate a character experience outline. In this way, the experience generation model can be guided to generate a preliminary character experience framework and key plots around the core settings and desired themes of the digital human. This makes the generated preliminary experience information more consistent with the digital human's settings and background, providing a solid foundation for subsequent refinement of the experience content.

[0065] It should be noted that the process of generating initial experience information may involve multiple rounds of human-computer interaction or collaboration between models to gradually enrich and refine the experience content. A pre-defined "prompt chain" guides the experience generation model through multi-stage, constrained generation. The prompts are designed following a universality principle, using placeholders to adapt to different digital human settings.

[0066] For example, the first prompt is: "Generate Character Experience Outline." The experience generation model generates preliminary experience information for the digital human based on the character definition and generation requirements. The character definition can include core identity, background, personality traits, and core skills. Generation requirements can include narrative themes, key plot points, and output format. For instance, the output format is a chronological list of key events, listing at least five core life stages or events that shape the character and drive the plot. Each event should be briefly described (no more than 50 words).

[0067] Step S4: Perform secondary processing on the primary experience information to obtain secondary experience information.

[0068] In this embodiment of the invention, secondary processing is used to verify and expand primary experience information. Primary experience information and key element information are input into the secondary model, and the secondary model is controlled by prompting engineering to perform secondary processing on the primary experience information. Finally, the secondary experience information output by the secondary model is obtained.

[0069] The secondary model includes at least one pre-trained large language model, and the prompting engineering includes a second prompt word, a third prompt word, and a fourth prompt word. The second prompt word is used to perform logical consistency verification on the primary experience information; the third prompt word is used to perform personalized enhancement on the primary experience information; and the fourth prompt word is used to perform diversity and innovative expansion on the primary experience information.

[0070] It should be understood that logical consistency verification checks whether the internal logic of the generated experience is self-consistent and whether the development of events conforms to the digital human's character settings and basic common sense. Personalization enhancement refines the behavior, dialogue, and decision-making in the experience based on the digital human's personality traits and skill settings, making them more in line with the character's traits. Diversity and innovative expansion introduces randomness or exploratory mechanisms to add more possibilities and innovative details to the experience while maintaining the core settings, avoiding overly homogenized generated content. Emotional and narrative pacing control adjusts the emotional tone and narrative rhythm of the experience based on the desired theme and effect, making the digital human's experience more engaging.

[0071] The secondary model, controlled by prompting engineering, performs secondary processing on the primary experience information, including: verifying whether the primary experience information contradicts the basic setting information and outputting the verification result, which includes setting contradiction and setting consistency; refining the behavior, dialogue, or decision in the primary experience information based on the basic setting information; and providing at least one selection path based on the decision points of the primary experience information to expand the primary experience information.

[0072] After secondary processing, the digital human's experience information aligns more closely with the basic settings and is more coherent and logical. The system can accurately identify and point out inconsistencies between the initial experience information and the digital human's basic settings, reinforcing the fit between experience and character design. The polishing process makes the digital human's behavior, dialogue, and decision-making more vivid, realistic, and layered. Providing alternative paths and expanding the initial experience information enriches the content of the digital human's experience, offering more possibilities and room for imagination for the digital human's interaction and story development.

[0073] For example, the second prompt is "Logical Consistency Analysis." The system verifies the initial experience information based on the character settings in the key element information. The analysis includes checking for contradictions in the events, timeline, and character behavior within the experience segments. It evaluates whether the character's behavior and decisions in the segments are consistent with their defined personality trait list. If not, please indicate and explain why. The output after verification begins with "Analysis Results" and lists all found logical or personality inconsistencies. If no issues are found, it outputs "Analysis Results: Logical and Personality Consistency Passed."

[0074] For example, the third prompt word is a personalized language polish. The system rewrites the original sentence according to the character's personality, making the sentence more personalized. The original sentence is "He saw the rain outside the window". If the digital human's personality is sentimental, then the polished sentence is "The rain outside the window seemed to be tapping on his heart as well, rippling in circles".

[0075] For example, the fourth prompt is to expand and refine key events. Based on the digital human's entry and exit settings, the system refines the events to be detailed. The requirements for refinement are: describe the complete process of the event from a third-person perspective. The narrative must explicitly or implicitly demonstrate one or more key character traits of the character. Describe the character's actions, psychological activities, and key dialogues. Ensure the plot development is logical and related to the character's core identity and core skills. Keep the word count around the expected limit.

[0076] Step S5: Format the secondary experience information according to the preset format and output the formatted secondary experience information.

[0077] In this embodiment of the invention, the preset format includes at least one of story text, time series, dialogue script, and structured data. By adjusting the format of the secondary experience information, the needs of different scenarios can be met.

[0078] For example, structured data includes fields such as timelines, event details, character relationship diagrams, and key dialogues, making it easy for programs to read and process. Long-form story text is used to describe complete biographical narratives. Dialogue scripts are used to drive interaction between the conversational digital human and the user.

[0079] This flexible format choice not only improves the usability of digital human experiences but also provides more possibilities for their application in the virtual world.

[0080] like Figure 2 As shown, this embodiment of the invention also provides a digital human character experience generation system for executing the above-described digital human character experience generation method. The digital human character experience generation system includes: an input parsing module 201, a material integration module 202, a core generation module 203, a secondary processing module 204, and a data output module 205.

[0081] The input parsing module 201 is configured to: acquire the basic setting information based on the digital human input by the user, and parse the basic setting information to obtain key element information.

[0082] The material integration module 202 is configured to: query relevant information in the knowledge material library that matches the key element information, including background information, plot fragments and narrative logic;

[0083] The core generation module 203 is configured to: input key element information and related information into the experience generation model, and obtain the primary experience information output by the experience generation model;

[0084] The secondary processing module 204 is configured to perform secondary processing on the primary experience information to obtain secondary experience information. The secondary processing is used to verify and expand the primary experience information.

[0085] The data output module 205 is configured to format the secondary experience information according to a preset format and output the formatted secondary experience information.

[0086] The following example, using the creation of a realistic digital human, illustrates the complete process of this invention.

[0087] The first step is to receive and parse the user's input of basic digital human settings and expected experience requirements.

[0088] User input information includes: "Name: Zhang San, Core identity: Post-90s independent documentary photographer, Personality traits: Introverted, observant, empathetic, resilient, Key skills: Photography, post-processing, basic wilderness survival skills, Desired experience theme: Growth, finding self-worth."

[0089] The parsed results are structured data, including:

[0090] {

[0091] "identity":"Documentary photographer",

[0092] "generation":"post-90s",

[0093] "personality": ["introverted", "observant", "empathetic", "resilient"]

[0094] "skills":["photography","post-processing","wilderness survival"],

[0095] Theme: "Growth, Finding Self-Worth"

[0096] "name":"Zhang San"

[0097] }

[0098] The second step is to retrieve and integrate relevant information from the knowledge resource database based on the analysis results.

[0099] The system retrieves relevant information about "documentary photographers" and integrates: the working styles of famous documentary photographers, knowledge about photographic equipment, important domestic and international photography awards, and the ethics of documentary photography.

[0100] The system retrieves narrative templates on the theme of "growth" and integrates them into a classic structure of "starting out, encountering setbacks, gaining inspiration, establishing direction, and achieving goals".

[0101] The third step is to generate a preliminary framework of the character's experiences and key plot points as initial experience information.

[0102] The system calls the first prompt word, fills in the parsed data, and requests the experience generation model to generate an experience outline.

[0103] The initial draft outline returned by the generative model is as follows:

[0104] During my university years: I hesitated between journalism and art photography, but ultimately chose the latter and began to explore documentary photography.

[0105] First independent project: After graduation, I traveled alone to a village that was about to disappear and used my camera to record the villagers' last days.

[0106] Encountering a professional ethical dilemma: When filming a sensitive social subject, one faces the difficult choice of "intervening as a recorder or observing from the sidelines".

[0107] Industry recognition: Her photography series on disappearing villages won a small national photography award, boosting her confidence.

[0108] A new journey: The decision was made to embark on a larger, long-term project to travel overseas and document the impact of climate change on Indigenous communities.

[0109] The fourth step is to perform secondary processing on the primary experience information to obtain secondary experience information.

[0110] The system selects point 2 in the outline, "First Independent Project," and uses the fourth prompt to expand upon it. The prompt will emphasize reflecting Zhang San's personality traits of being "introverted but observant" and "empathetic."

[0111] The generated story clips will describe in detail how she quietly integrates into the village without disturbing the villagers, but captures the details of their lives, the light and shadow and the emotional expressions through her lens, and how she puts down her camera and listens quietly when an old man tells his story, demonstrating her empathy.

[0112] The system invokes a second prompt to perform a consistency check on the generated story fragments. For example, it checks whether her interactions with the villagers in the story conform to the "introverted" setting. If the model generates a scene where she engages in lengthy, rambling conversations, the validation module will mark it as "inconsistent personality" and trigger a regeneration or modification.

[0113] The description "she took a photo" can be refined by using a third cue word, changing it to "She held her breath, captured a fleeting moment of loneliness on the old man's deeply lined face, and gently but decisively pressed the shutter." This makes it more in line with the setting of "keen observation skills".

[0114] Step 5: Output the secondary experience information.

[0115] The system will integrate the refined and verified complete experience into a JSON file containing a character profile, detailed experience story, timeline of key events, and relationship diagram.

[0116] {

[0117] "character_name":"Zhang San",

[0118] "summary": "A resilient and empathetic documentary photographer born in the 1990s..."

[0119] "timeline":[

[0120] {

[0121] "time":"University period",

[0122] "event_key":"Establish direction",

[0123] "description": "Hesitating between journalism and art photography, I ultimately chose the latter and began exploring documentary photography."

[0124] },

[0125] {

[0126] "time":"One year after graduation",

[0127] "event_key":"First Independent Project",

[0128] "title":"The Last Village",

[0129] "narrative": "Zhang San, alone with her backpack and camera, arrived at a village nestled deep in the mountains, slated to be submerged during reservoir construction. Initially, her introverted nature created a distance between her and the villagers. But instead of rushing to photograph, she spent a week quietly observing, helping Grandpa Zhang feed the chickens, and listening to Grandma Li tell stories of the past. Her keen observation allowed her to capture many previously unknown details… Ultimately, her series 'The Last Nostalgia' truthfully and profoundly records the village's disappearance, and also allows her to find her value as a recorder."

[0130] "personality_reflection": ["Introverted", "Observant", "Empathic"]

[0131] }

[0132] / / ...other events ]

[0134] }

[0135] Through the above steps, this invention successfully generated a realistic character experience for the digital human "Zhang San" that is consistent with his persona, logically coherent, rich in content, and deeply personalized. This experience can be directly used to drive the digital human's behavior, dialogue, and story presentation in the virtual world, greatly enhancing the digital human's realism and appeal.

[0136] As can be seen from the above technical solutions, this invention discloses a method and system for generating digital human character experiences. By utilizing the generation capabilities of large models, it significantly reduces the workload of manually compiling character experiences, greatly improves the efficiency of experience generation, and generates highly personalized character experiences based on user input, so that each digital human has a unique life experience, enhancing the uniqueness and realism of the digital human. Furthermore, through secondary processing, it ensures that the generated experiences are logically self-consistent and consistent with the core character settings, improving the credibility of the experiences, thereby enhancing the realism, attractiveness, and application value of the digital human.

[0137] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating the experiences of a digital human, characterized in that, include: Obtain basic settings information based on digital human input from the user, and parse the basic settings information to obtain key element information; The relevant information in the knowledge resource library that matches the key element information includes background information, plot fragments, and narrative logic. The key element information and the relevant information are input into the experience generation model, and the primary experience information output by the experience generation model is obtained. The primary experience information is processed to obtain secondary experience information, which is used to verify and expand the primary experience information. The secondary experience information is formatted according to a preset format, and the formatted secondary experience information is output.

2. The method for generating the character experience of a digital human according to claim 1, characterized in that, The basic setting information includes at least one of the following: identity tags, personality traits, key skills, desired themes, and specific event nodes.

3. The method for generating the character experience of a digital human according to claim 2, characterized in that, The parsing of the basic setting information yields key element information, including: The basic setting information is preprocessed, and the preprocessing includes: cleaning, splitting, and standardization; The preprocessed basic setting information is structured and parsed using a large language model to obtain the key element information.

4. The method for generating the character experience of a digital human according to claim 1, characterized in that, The knowledge resource library includes at least one of the following: general world knowledge, domain-specific knowledge, narrative structure templates, plot elements, and emotional patterns.

5. The method for generating the character experience of a digital human according to claim 4, characterized in that, The relevant information in the knowledge resource database that matches the key element information includes: Retrieve multidimensional information corresponding to the key element information from the knowledge resource base; By integrating the multidimensional information, the relevant information is obtained.

6. The method for generating the character experience of a digital human according to claim 2, characterized in that, The experience generation model includes at least one pre-trained large language model; The method for generating the digital human's experience also includes: The experience generation model is controlled by a preset prompting process to output the primary experience information, which is used to characterize the digital human's experience framework and key plots. The prompting process includes a first prompt word, which is used to instruct the experience generation model to generate a character experience outline.

7. The method for generating the character experience of a digital human according to claim 6, characterized in that, The secondary processing of the primary experience information to obtain secondary experience information includes: The primary experience information and the key element information are input into the secondary model, which includes at least one pre-trained large language model. The prompting engineering controls the secondary model to perform secondary processing on the primary experience information; the prompting engineering includes a second prompt word, a third prompt word, and a fourth prompt word, wherein the second prompt word is used to perform logical consistency verification on the primary experience information; the third prompt word is used to perform personalized enhancement on the primary experience information; and the fourth prompt word is used to perform diversity and innovative expansion on the primary experience information. Obtain the secondary experience information output by the secondary model.

8. The method for generating the character experience of a digital human according to claim 7, characterized in that, The process of using the prompting engineering control to perform secondary processing on the primary experience information in the secondary model includes: Verify whether the initial experience information contradicts the basic setting information, and output the verification result, which includes setting contradiction and setting consistency; Refine the behaviors, dialogues, or decisions in the primary experience information based on the aforementioned basic setting information; Based on the decision points of the primary experience information, at least one selection path is provided to expand the primary experience information.

9. The method for generating task experiences for a digital human according to claim 1, characterized in that, The preset format includes at least one of story text, time series, dialogue script, and structured data.

10. A system for generating the character experience of a digital human, used to execute the method for generating the character experience of a digital human as described in any one of claims 1 to 9, characterized in that, include: The input parsing module is configured to: acquire basic setting information based on digital human input by the user, and parse the basic setting information to obtain key element information; The material integration module is configured to: query relevant information in the knowledge material library that matches the key element information, the relevant information including background information, plot fragments and narrative logic; The core generation module is configured to: input the key element information and the relevant information into the experience generation model, and obtain the primary experience information output by the experience generation model; The secondary processing module is configured to perform secondary processing on the primary experience information to obtain secondary experience information, wherein the secondary processing is used to verify and expand the primary experience information; The data output module is configured to: format the secondary experience information according to a preset format, and output the formatted secondary experience information.

Images