An AI-based multi-modal game generation method and generation system

Abstract

The application relates to an AI-based multi-modal game generation method and a generation system, obtains associated information groups for generating games, analyzes and generates a structured initial game design list, revises the initial game design list based on requirements, calls multi-modal agent groups to generate corresponding game component images under each classification in the revised list, repeatedly cross-checks the multi-modal agent groups, coordinates until the requirements are met, fuses all generated game component images according to the associated information groups, and generates games; the generation system comprises a list generation module, a multi-modal generation module and a fusion output module. The application reduces the requirement of game design on the professionalism of users, expands the applicable population, simultaneously reduces the difficulty of game design for developers, improves the efficiency of game design, and effectively reduces the game design cost; the generated game product can be commercially used without repeated modification, the game is systematically fused in the whole process, the operation is simple, and the game is easy to use.

Description

Technical Field

[0001] This invention relates to the technical fields of card games, board games, or roulette games; indoor games using small moving objects; video games; and games not included in other categories. In particular, it relates to an AI-based multimodal game generation method and system in the field of artificial intelligence control. Background Technology

[0002] With the large-scale development of the gaming industry, game development has evolved into a complex process encompassing early planning, art design, technical development, sound effects and music, testing and optimization, and release, operation, and maintenance. Currently, the industry demands higher levels of richness, personalization, and speed of updates for game content. Especially with the "Games as a Service" (GaaS) model becoming the mainstream trend, continuously producing high-quality content has become a core element for maintaining user engagement. To address these challenges, the industry has begun to explore incorporating artificial intelligence (AI) technology into the development process. Existing AI applications primarily focus on assisting specific, localized stages. In terms of content generation, models such as Generative Adversarial Networks (GAN), Variational Autoencoders (VAE), and diffusion models are used to achieve the automated generation of digital assets such as scene terrain, character concept art, and sound effects. For example, Stable Diffusion supports fine customization of art resources through the ControlNet tool, while DALL-E3 can generate coherent scene concept maps based on complex text prompts. In terms of efficiency optimization, some manufacturers have improved the efficiency of 2D image generation (60%~80%) and increased the proportion of game video delivery to over 70% through AI tools, and have also applied AI to translation calibration, customer service response and other processes. However, while AI technology has already played a role in some aspects of game development, its ability to generate professional and systematically adaptable content remains insufficient. Currently, most AI tools produce generic materials lacking specificity, failing to meet the unique style, gameplay logic, and other professional requirements of specific games, and also lacking direct commercial value. These materials generally suffer from severe homogenization, rough details, and a disconnect from the core game design. Incorporating them into the actual development process requires significant manual intervention for targeted modifications, logic calibration, and quality optimization, which not only greatly increases development costs but may also impact development efficiency due to repeated adjustments, failing to truly address the core pain points of game development. Summary of the Invention

[0003] This invention solves the problems existing in the prior art and provides an AI-based multimodal game generation method and system.

[0004] The technical solution adopted in this invention is an AI-based multimodal game generation method, wherein the method obtains a group of related information for generating games, parses and generates a structured initial game design list, and revises the initial game design list based on requirements; The multimodal intelligent agent group generates and revises the game component images corresponding to each category in the list; the multimodal intelligent agent group repeatedly cross-validates and coordinates until it meets the preset; All generated game component images are merged according to the associated information group to generate the game.

[0005] Preferably, the associated information group includes a game template identifier and game style requirements, which are parsed by the first AI agent to obtain the configuration information of the corresponding game template; the game template includes the coordinate information and configuration parameters of the corresponding game components.

[0006] Preferably, the game template is a preset or custom-built template; Input template configuration information through a preset configuration channel to generate a configuration file containing the coordinate information and configuration parameters of game components and a template file without materials. The second AI agent generates exemplary game component images with the same style as the game theme based on the art style in the template configuration information and stores them together.

[0007] Preferably, the multimodal intelligent agent group includes an interface intelligent agent, a map intelligent agent, a character intelligent agent, an icon intelligent agent, a special effects intelligent agent, a logo intelligent agent, and a verification intelligent agent; the verification intelligent agent is used to parse the numerical and non-numerical constraints in the game design list and convert them into generation parameters for use by other intelligent agents, and to perform numerical and non-numerical compliance verification on the outputs of other intelligent agents.

[0008] Preferably, cross-validation is used to detect whether there is information conflict between game component images generated by different agents. If there is, the generation parameters of at least one agent are adjusted based on preset conflict resolution rules. The adjusted agent regenerates or adjusts the corresponding game component images until the conflict is resolved.

[0009] Preferably, the game components include characters, interface, icons, special effects, map, logos, and numerical values.

[0010] Preferably, when replacing the game component image with the corresponding position in the game template, the image size is detected to match the size of the preset position in the template; if they do not match, adjustments are made. If an art style conflict is detected between different game component images, then based on the game style requirements or preset priority rules, style regeneration or style transfer processing is triggered for at least one conflicting game component image.

[0011] An AI-based multimodal game generation system includes: The list generation module is used to generate game-related information groups, parse and generate a structured initial game design list, and revise the initial game design list based on requirements; A multimodal generation module is used to generate game component images corresponding to each item in the revised game design list by calling a multimodal intelligent agent group; the multimodal intelligent agent group includes several intelligent agents, which are dedicated to processing specific types of game components, and a cross-validation unit is provided in conjunction with the multimodal intelligent agent group; The fusion output module is used to merge the generated game component images according to the guidance of the associated information group to generate the finished game.

[0012] Preferably, the multimodal intelligent agent group includes an interface intelligent agent, a map intelligent agent, a character intelligent agent, an icon intelligent agent, a special effects intelligent agent, a logo intelligent agent, and a verification intelligent agent; a collaborative monitoring platform is provided in conjunction with the multimodal intelligent agent group to transmit parameters and data between the intelligent agents and to collaboratively monitor the results and parameters generated by each intelligent agent; the verification intelligent agent obtains the generation requirements and output results of other intelligent agents through the collaborative monitoring platform and performs parameter conversion and compliance verification.

[0013] Preferably, it further includes a template building module, the template building module comprising: The configuration interface is used to receive template configuration information input by the user. A template file generator is used to generate a configuration file and a template file without materials based on the template configuration information. A template example generator is used to generate example components with a style consistent with the template configuration information using a second AI agent and associate them with the template file.

[0014] This invention relates to an AI-based multimodal game generation method and system. The method involves acquiring a set of associated information for game generation, parsing and generating a structured initial game design list, revising the initial game design list based on requirements, calling a multimodal intelligent agent group to generate and revise game component images corresponding to each category in the revised list, repeatedly cross-validating and coordinating the multimodal intelligent agent group until a preset condition is met, and fusing all generated game component images according to the associated information set to generate the game. The generation system includes a list generation module, a multimodal generation module, and a fusion output module.

[0015] The beneficial effects of this invention are as follows: (1) This invention reduces the professional requirements of users for game design, expands the applicable population, and at the same time reduces the difficulty for developers to design games; (2) This invention allows users to build game templates themselves and then generate finished games with one click based on the game templates, which improves the efficiency of game design. The game materials designed by the method are both targeted and consistent in style. The generated finished games can be commercially launched without repeated modifications, which effectively reduces the cost of game design. (3) By integrating various game component images with game templates, a customized finished game can be formed. The operation is simple and achieves the systematic integration of the entire game process, effectively avoiding the problem of fragmented processes and low development efficiency caused by the mutual coordination of data in each stage. At the same time, the integration of art materials and templates can be achieved simply by replacing art materials through the associated interface. The operation is simple and easy to learn. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the specific implementation process of the present invention; Figure 3 A flowchart for configuring a game template for this invention; Figure 4 This is a schematic diagram illustrating the operation of the multimodal intelligent agent group in this invention; Figure 5 This is a schematic diagram of the system structure of the present invention; Figure 6 This describes the application workflow of the intelligent agent in this invention. Detailed Implementation

[0017] 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.

[0018] This invention relates to an AI-based multimodal game generation method, which constructs a cascading transformation path of "requirements-list-components-finished product". From extracting user requirements to AI parsing them into a structured design list, then multiple agents parse the list and generate components, and finally the components are integrated with templates to form the finished game. This method gradually transforms highly unstructured game ideas into structured intermediate products and the final executable game, lowering the professional threshold while ensuring the controllability and operability of the process.

[0019] The method includes the following steps: S1 acquires the associated information group used to generate the game, parses and generates a structured initial game design list, and revises the initial game design list based on requirements; S2 calls the multimodal intelligent agent group to generate and revise the game component images corresponding to each category in the list; the multimodal intelligent agent group repeatedly cross-validates and coordinates until it meets the preset; S3 merges all generated game component images according to the associated information group to generate the game.

[0020] The method will be described below with reference to specific implementation methods.

[0021] S1 acquires the associated information group used to generate the game, parses and generates a structured initial game design list, and revises the initial game design list based on requirements; In this invention, a precise guidance and fusion mechanism based on templates without material is established. By introducing game templates without art materials but containing only coordinates and configuration information, precise spatial constraints are provided for AI generation, making the game generation process more targeted, significantly reducing the uncertainty of the generation results, and improving usability.

[0022] Specifically, the associated information group includes a game template identifier and game style requirements, which are parsed by the first AI agent to obtain the configuration information of the corresponding game template; the game template includes the coordinate information and configuration parameters of the corresponding game components, and most importantly, there are no art assets in the game template; After the user selects a game template and describes the game style requirements, the corresponding game template identifier and game style requirements are input into the first AI agent. After the first AI agent parses and obtains the coordinate information and configuration parameters of the game components, it "understands" the game style requirements and generates a design list that conforms to the requirements logic, including game type, game values, game background, game focus, and background information of each game component. Through the game list, all the key elements of the finished game can be clearly understood. If the current game list does not meet the requirements, the list can be modified as a whole or by modifying individual elements. This can be done by changing the design style or manually editing the list content. If it is necessary to modify the relevant content of the game components, it can be done by manually modifying (with the first AI agent polishing the text) or by having the first AI agent actively modify it for optimization.

[0023] Furthermore, the game template can be preset or custom-built; When building a game template, users configure the template through the game's preset channel, including the following steps: S1.1 Input template configuration information through the preset configuration channel to generate a configuration file containing the coordinate information and configuration parameters of game components and a template file without materials; the configuration standards include numerical configuration standards and art material configuration standards. The art material configuration standards include: the configuration of interface, map, character, icon, special effect and logo; Then, the generated game configuration file and template file are obtained, and the current type, layout, name, game content and preset source are configured. The game content includes the art style, game theme and game rules. S1.2 The second AI agent generates an exemplary game component image that matches the art style and game theme in the template configuration information and stores it in association. This association storage includes association with the template file and association between components.

[0024] The application of the game template proposed in this invention simplifies the difficulty of the game planning stage and reduces the professional ability requirements of users for game design.

[0025] S2 calls the multimodal intelligent agent group to generate and revise the game component images corresponding to each category in the list; the multimodal intelligent agent group repeatedly cross-validates and coordinates until it meets the preset; In this invention, a multi-agent architecture that is both distributed and collaborative is adopted. By constructing a "multimodal agent group" that includes multiple agents with different modalities and purposes, each agent focuses on the generation of game components in a specific direction, ensuring the professionalism of individual agents and the generation quality of corresponding game components. During the process, by verifying agents as the core of collaboration and cross-verification process, and in conjunction with the information of the associated information group, the consistency of the generation parameters and results of each component is achieved, thus solving the common problems of "style fragmentation" and "setting conflict" in multimodal generation.

[0026] The multimodal intelligent agent group includes an interface intelligent agent, a map intelligent agent, a character intelligent agent, an icon intelligent agent, a special effects intelligent agent, a logo intelligent agent, and a verification intelligent agent; the verification intelligent agent is used to parse the numerical and non-numerical constraints in the game design list and convert them into generation parameters for use by other intelligent agents, and to perform numerical and non-numerical compliance verification on the outputs of other intelligent agents. Each intelligent agent focuses on a specific area of ​​game image generation, while achieving deep information fusion through cross-module collaboration mechanisms; The following provides a detailed explanation of each intelligent agent.

[0027] 2.1 Interface Intelligent Agent The interface intelligence agent is used to handle UI-related requirements. It can parse the interface layout in the planning list, such as generating an interface framework that conforms to operating habits based on the text description "the main interface should include character avatars, backpack entrances, and task tracking bars, distributed vertically".

[0028] 2.2 Map Agent The map agent is responsible for interpreting the setting information of the map scene, such as "the beginner village map should include forest, river and wooden house elements, and the overall color scheme should be green". Combined with terrain numerical parameters, such as "the map size is 1000×800 pixels and the river width accounts for 15%", it generates scene images that match the world view.

[0029] 2.3 Role-based Intelligent Agents The character AI agent identifies the character's skin, appearance details, etc., extracts information such as the character's profession, personality, and clothing characteristics from the planning list, and generates skin designs that match the character's positioning, such as the profession being "assassin", the personality being "cold", and the clothing characteristics being "black bodysuit + red cape".

[0030] 2.4 Icon-based Intelligent Agents The icon intelligence agent is used to generate skill icons and item icons, and to parse the icon function description and size constraints, such as "the healing skill icon should reflect a sense of healing and be based on a cross symbol" and "the icon should be 64×64 pixels with a rounded corner radius of 5px", to ensure that the icons are intuitive and conform to the specifications.

[0031] 2.5 Special Effects Intelligent Agent The special effects agent processes the visual requirements of skill effects and scene effects, such as "the fireball effect should include flame spray and explosion light effects, lasting 0.5 seconds", and generates a dynamic effect sequence by combining the effect frame rate, such as "2 frames / second".

[0032] 2.6 Flag Intelligent Agent The logo intelligence agent designs game logos, faction symbols, etc., analyzes the core elements and style requirements of the logo, such as "the logo must contain a dragon pattern and the letter 'K'" and "retro metal style", and generates a logo image with high recognizability.

[0033] 2.7 Verify the intelligent agent As the core of collaboration, the verification agent is responsible for processing all numerical and non-numerical constraints in the planning list. Numerical constraints include combat experience, BOSS strength, monster strength, protagonist growth pace (fast in the early stage and difficult in the late stage), etc., and converting them into generation parameters that can be recognized by other agents. At the same time, it verifies whether the output results of each agent meet the numerical requirements. Cross-validation is used to detect whether there are information conflicts (non-numerical constraints) between game component images generated by different agents. If such conflicts exist, such as the scene color tone generated by the map agent not matching the logo style of the logo agent, the generation parameters of at least one agent are adjusted based on preset conflict resolution rules. Generally, preset conflict resolution rules include unifying game styles and prioritizing scene color tone over logo style. During cross-validation, the associated information group is used as the main body. When other agents get stuck in deadlock, the associated information group is given the decision-making power to manage information conflicts and the verification is terminated.

[0034] The adjusted agent regenerates or adjusts the corresponding game component images until the conflict is resolved.

[0035] The final generated game components include characters, interface, icons, special effects, maps, logos, and numerical values.

[0036] In this invention, the application of intelligent agents can be implemented through conventional methods. Generally, multiple readily available AI components, including but not limited to large models, LoRA adapters, and post-processing tools, are connected through an intelligent agent workflow. For example, when a user inputs text or an image, the system selects a corresponding large model from a set of known large models for text polishing or image parsing. Then, the model enters the workflow of the corresponding intelligent agent for pre-processing steps such as noise reduction and style determination using LoRA. Finally, a common open-source model, such as Flax, is used to generate the specific content. The specific content is then subjected to post-processing workflows such as frame extraction, frame interpolation, and image matting to generate the final result. If the agent requires training in an application, it is generally trained based on the post-training mode of the open-source model; this is something that is easy for those skilled in the art to understand, and they can carry it out through conventional large model implementation methods.

[0037] S3 merges all generated game component images according to the associated information group to generate the game.

[0038] Generally, based on the newly generated game component images, the corresponding positions of each component in the game template are replaced with the new component images through the associated interface. After all the game component images are replaced, a finished game that meets the user's needs can be generated.

[0039] This invention also introduces a dynamic adaptation and conflict correction mechanism in the material and template fusion stage, which can automatically handle technical problems such as size mismatch and style conflict, such as calling the verification agent to perform parameter conversion, forming an automated closed loop from design, generation to final integration, enhancing the practicality and robustness of the method, and ensuring the overall consistency after splicing and fusion.

[0040] Specifically, when replacing the game component image with the corresponding position in the game template, the image size is checked against the size of the preset position in the template. If they do not match (e.g., the icon generated by the icon agent is 60×60 pixels, while the template position requires 50×50 pixels), adjustments are made. In general, the replacement engine calls the verification agent to perform parameter conversion and automatically scales the icon to the matching size while maintaining the image clarity. If an art style conflict is detected between different game component images, the game components are generally configured with art style parameters, and the conflict can be detected by checking whether the parameters are consistent. For example, if the parameters of the sci-fi style character generated by the character agent and the ancient style map generated by the map agent do not match, then based on the game style requirements or preset priority rules, style regeneration or style transfer processing of at least one conflicting game component image will be triggered.

[0041] This invention also relates to an AI-based multimodal game generation system, comprising: (1) Inventory generation module The list generation module is used to generate game-related information groups, parse and generate a structured initial game design list, and revise the initial game design list based on requirements; It interacts with users through a UI interface, obtains the game template selected by the user and the natural language style requirements entered by the user. The module calls the first AI agent, such as a large language model LLM, which is trained or prompted to understand knowledge in the game design domain. Combined with the configuration file of the selected template, it parses, expands and structures the fuzzy style requirements, and generates an initial game design list. The list information includes game metadata, art style requirements, component-level design summary, etc. After presenting the initial game design list to the user, an editing interface is provided, allowing the user to revise the list in whole or in part. The revised list serves as input for the multimodal generation module.

[0042] (2) Multimodal generation module A multimodal generation module is used to generate game component images corresponding to each item in the revised game design list by calling a multimodal intelligent agent group; the multimodal intelligent agent group includes several intelligent agents, which are dedicated to processing specific types of game components, and a cross-validation unit is provided in conjunction with the multimodal intelligent agent group; The multimodal intelligent agent group consists of a series of AI agents used to generate different game components, each of which is a specialized model optimized or trained for a specific domain. Specifically, the multimodal intelligent agent group includes an interface agent, a map agent, a character agent, an icon agent, a special effects agent, a logo agent, and a verification agent; The UI agent inputs textual descriptions and layout constraints related to the UI from the input list, and outputs interface wireframes or mockups that conform to human-computer interaction standards; the map agent inputs scene descriptions and terrain parameters, and outputs an overall scene image that conforms to the world view and stitchable map tiles; the character agent inputs character settings (profession, personality, clothing), and outputs skin or 3D model textures; the icon agent inputs icon function descriptions and size specifications, and outputs clear and easily recognizable skill, prop, and status icons; the special effects agent inputs special effects descriptions and dynamic parameters (frame count, duration), and outputs sequence frame images or particle effect parameters; the logo agent inputs the core elements and style of the logo, and outputs a recognizable LOGO or badge pattern; the verification agent, as the collaborative core, does not directly generate images, but instead analyzes all numerical and non-numerical balance requirements in the input list, and performs parameter transformation and compliance verification.

[0043] The multimodal intelligent agent group is equipped with a collaborative monitoring platform, which is used to transmit parameters and data between the agents and collaboratively monitor the results and parameters generated by each agent. The verification agent obtains the generation requirements and output results of other agents through the collaborative monitoring platform and performs parameter conversion and compliance verification.

[0044] The collaborative monitoring platform provides a shared data exchange channel for each intelligent agent. After the list is parsed, the relevant tasks and parameters are distributed to the corresponding specialized intelligent agents through the collaborative monitoring platform. During the process, the collaborative monitoring platform monitors whether the results generated by each intelligent agent are successful and whether each parameter meets the requirements, thereby realizing collaboration between multiple intelligent agents. It also proposes a cross-validation unit that runs during or after generation to monitor conflicts between the outputs of each agent. When a conflict does exist, the cross-validation unit calls a preset conflict resolution rule base and instructs the relevant agents to regenerate or adjust the system via the cooperative bus.

[0045] (3) Fusion output module The fusion output module is used to merge the generated game component images according to the guidance of the associated information group to generate the finished game.

[0046] In this invention, the fusion output module reads the configuration file through the engine to obtain the coordinates and attributes of each game component, and replaces the corresponding placeholders in the template file with the game component images output by the multimodal generation module through the associated interface; during the replacement process, a dynamic adaptation submodule implements dynamic adaptation and conflict correction, including: Size adaptation: If the size of the icon generated by the icon agent does not match the reserved position in the template configuration, the dynamic adaptation submodule will automatically call the image scaling algorithm and use the verification agent to ensure that the visual clarity after scaling meets the preset quality standards. Final style harmonization: In the final integration stage, if style inconsistencies are still detected, a lightweight style transfer network is invoked to fine-tune individual components and ensure overall visual consistency.

[0047] After all components are replaced and adapted, the merge output module packages all resources and outputs a complete, compileable, executable game project or a game application package that can be directly experienced.

[0048] The generation system also includes (4) a template building module; The template building module provides users with a channel to create personalized game templates. Its core function is to generate a game template without any assets, containing only the logic and structural framework. The template building module includes: The configuration interface provides a graphical user interface for receiving template configuration information input by the user. The configuration information follows preset numerical configuration standards and art material configuration standards. The template file generator is used to generate a configuration file and a template file without materials based on the template configuration information. The configuration file records all coordinate information, hierarchical relationships, initial numerical parameters and component type identifiers of the game components in a structured data format such as JSON or XML. The template file without materials is the executable framework or project file of the game, which is configured with the location of the game components. The template example generator is used to generate example components with the same style as the template configuration information using a second AI agent (such as a finely tuned text-to-image model) and associate them with the template file. During the process, the second AI agent generates a set of exemplary component images with a unified style according to the user-configured "art style" and "game theme". The example images are not bound to the template logic, but only serve as a style preview and reference for the subsequent generation process. They are associated with the template file and stored in the template library.

[0049] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0050] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0051] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0052] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0053] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0054] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A multimodal game generation method based on AI, characterized in that: The method obtains a group of associated information for generating the game, parses and generates a structured initial game design list, and revises the initial game design list based on requirements; Call the multimodal intelligent agent group to generate and revise the game component images corresponding to each category in the list; The multimodal intelligent agent group repeatedly cross-validates and coordinates until it meets the preset requirements; All generated game component images are merged according to the associated information group to generate the game.

2. The AI-based multimodal game generation method according to claim 1, characterized in that: The associated information group includes game template identifiers and game style requirements. After being parsed by the first AI agent, the configuration information of the corresponding game template is obtained. The game template includes the coordinate information and configuration parameters of the corresponding game components.

3. The AI-based multimodal game generation method according to claim 2, characterized in that: The game template can be preset or custom-built; Input template configuration information through a preset configuration channel to generate a configuration file containing the coordinate information and configuration parameters of game components and a template file without materials. The second AI agent generates exemplary game component images with the same style as the game theme based on the art style in the template configuration information and stores them together.

4. The AI-based multimodal game generation method according to claim 1, characterized in that: The multimodal intelligent agent group includes an interface intelligent agent, a map intelligent agent, a role intelligent agent, an icon intelligent agent, a special effects intelligent agent, a logo intelligent agent, and a verification intelligent agent; The verification agent is used to parse the numerical and non-numerical constraints in the game design list and convert them into generation parameters for use by other agents. The outputs of other agents are then verified for numerical and non-numerical compliance.

5. The AI-based multimodal game generation method according to claim 4, characterized in that: Cross-validation is used to detect whether there are information conflicts between game component images generated by different agents. If there are, the generation parameters of at least one agent are adjusted based on the preset conflict resolution rules. The adjusted agent regenerates or adjusts the corresponding game component images until the conflict is resolved.

6. The AI-based multimodal game generation method according to claim 4, characterized in that: The game components include characters, interface, icons, special effects, maps, logos, and numerical values.

7. The AI-based multimodal game generation method according to claim 1, characterized in that: When replacing the game component image with the corresponding position in the game template, the image size is checked against the size of the preset position in the template. If they do not match, adjustments are made. If an art style conflict is detected between different game component images, then based on the game style requirements or preset priority rules, style regeneration or style transfer processing is triggered for at least one conflicting game component image.

8. An AI-based multimodal game generation system, characterized in that: include: The list generation module is used to generate game-related information groups, parse and generate a structured initial game design list, and revise the initial game design list based on requirements; The multimodal generation module is used to generate game component images corresponding to each item in the revised game design list by calling a multimodal intelligent agent group. The multimodal intelligent agent group includes several intelligent agents, which are dedicated to processing specific types of game components, and a cross-validation unit is provided in conjunction with the multimodal intelligent agent group; The fusion output module is used to merge the generated game component images according to the guidance of the associated information group to generate the finished game.

9. The AI-based multimodal game generation system according to claim 8, characterized in that: The multimodal intelligent agent group includes an interface intelligent agent, a map intelligent agent, a character intelligent agent, an icon intelligent agent, a special effects intelligent agent, a logo intelligent agent, and a verification intelligent agent. A collaborative monitoring platform is provided in conjunction with the multimodal intelligent agent group to transmit parameters and data between the agents and to collaboratively monitor the results and parameters generated by each agent. The verification intelligent agent obtains the generation requirements and output results of other agents through the collaborative monitoring platform and performs parameter conversion and compliance verification.

10. The AI-based multimodal game generation system according to claim 8, characterized in that: It also includes a template building module, which includes: The configuration interface is used to receive template configuration information input by the user; A template file generator is used to generate a configuration file and a template file without materials based on the template configuration information. A template example generator is used to generate example components with a style consistent with the template configuration information using a second AI agent and associate them with the template file.

Images