May 20 News.Tencent HunyuanThe team just announced: Tencent MixedGame Visual Generation Platform(hereinafter referred to as "Hybrid")game") was officially released. This is the first industrial-grade AIGC A game content production engine that dramatically optimizes the game asset generation and game production process.
According to the official introduction, Tencent mixed yuan will Hunyuan-Game position the first industrial-grade game content asset generation for the creation of the AIGC series of large models.
The technical report shows that it covers a complete game image and video modeling capability matrix, including four major game image models and five major video models. The four major image generation models include the Game Raw Image Large Model Base, the Special Effects Generation Model, the Transparency Generation Model, and the Character Original Drawing Generation Model.
According to reports, the capabilities that have been put online at this stage of the hybrid game include AI art pipeline, real-time canvas, AI 2D art generation, professional character multi-view generation and other core capabilities, which are focused on the conceptual design of the game's art.
Among them, the AI art pipeline capability seamlessly integrates multiple AI tools, allowing you to "complete the entire workflow of getting inspiration -> sketching character concepts -> generating standard character multi-views" on a single page without any need to jump, realizing a zero-transfer from "inspiration" to "delivery". Delivery" from "inspiration" to "delivery", realizing zero jump.
For example, if a designer enters a text such as "a cute anime girl with thick paint style", AI can instantly generate a character inspiration gallery and complete hand-drawn sketch rendering, standard three-view generation and 360° dynamic demo video output on the same page, bidding farewell to the fragmented process of multi-tool switching.
Real-Time Canvas is a real-time AI capability that allows designers to quickly generate and instantly correct conceptual designs through hand-drawn sketches, improving creative efficiency "while drawing". The designer draws rough lines, sketches schematics, and adds descriptions with prompts, and the AI generates and presents the results in about a second, rendering the details in sync.
Designers can adjust strokes, character postures, scene lighting and shadows at any time and get instant feedback, allowing users to focus on creativity itself and realize the immersive creative experience of "changing while drawing", which realizes a higher degree of freedom of creativity and significantly improves productivity at the same time. At the same time, based on the world's first real-time generation engine developed by Tencent Hybrid, it avoids the long waiting time in the traditional drawing era.
AI 2D Art Inspiration provides AI text-to-diagram / diagram-to-diagram capabilities, behind which is a game-specific 2D art graphics model, which is able to understand a variety of game terminology, concepts, and style descriptions, realizing a more accurate reproduction of conceptual creativity in the field of gaming.
For example, if you type in "Warcraft style silver shield with runes around it", you can generate material that can be used directly for 3D modeling; uploading a weapon sketch will generate a series of variants with "blue blade and jeweled grips"; and it also supports the generation of transparent channel maps. Generate.
According to the official introduction, the game-specific model is based on millions of game and animation data training pendant model, with native Chinese Prompt in-depth parsing capabilities, can accurately restore the "thick coating, celluloid, cyberpunk" and other terminology, support for a variety of mainstream game styles and themes, such as realism, cartoon, oriental myths, fairy fantasy and so on.
In the traditional game art design process, game art designers need to draw a three-view of the character for the character after producing the original drawing of the character.
With Hybrid's game visualization capabilities, game art designers only need to upload a single character drawing, and Hybrid's AI can generate industry standard three-views and 360° rotated videos for subsequent 3D modeling in a single click.
If designers are not satisfied with a generated detail, they can manually modify it. Character multi-view generation time is compressed from the traditional 12 hours to a few minutes to 30 minutes, and the consistency of the main body reaches 99%. At present, this ability has been applied to a number of headline game projects, helping the art team to improve efficiency by 300%.
-
With experience URL:
https://hunyuan.tencent.com/game/ -
Technical report:
https://cdnportal.hunyuan.tencent.com/game/assets/pdf/Hunyuan_Game.pdf