VaM2 AI Empowers Game Animation Creation – Let's Discuss the Necessity of Open APIs!

[Jack Ma]

Hello, all players and esteemed developers! As a beginner deeply fascinated by hyper-realistic characters and exquisite graphics in games, I've been truly discouraged and stopped trying several times recently when attempting to create my own game animations due to the complex operations . Today, I want to discuss with everyone: If games were to open up standardized API interfaces to allow AI to assist with animation creation, would this be a win-win direction?​

My Core Feeling: It’s Too Hard for Beginners to Make Animations!
I wonder if there are other employed players like me? To create a satisfying game animation often requires spending several days studying the operations. The learning cost and time commitment are ridiculously high. I have so many ideas I want to realize, but getting stuck on the complicated animation production steps is truly frustrating...
My Idea: Leverage AI, Open Standardized API Interfaces
AI technology is developing so fast now, especially in the fields of AIGC (AI-Generated Content) and motion generation, where it has already achieved many stunning effects. So I was thinking, could game developers consider opening up and encapsulating some core functional modules to provide a set of standardized API interfaces specifically for AI?
I want to extend this by mentioning the "MCP Server Interface" I followed before: it doesn't necessarily have to strictly adhere to a specific technical standard, but its core idea – "providing external applications with secure, standardized access to data and functionalities" – is what I think is particularly crucial.
Specifically, I think the encapsulated interfaces could include these capabilities (welcome everyone to add more!):

• Action Sequence Generation/Import API: Allows AI to directly generate or import details like skeletal animation data, facial expression data, and even cloth simulation parameters.
• Character/Environment Control API: Supports both high-level simple commands like "walk to the chair and sit down" and fine-grained operations like directly controlling joint rotations.
• State and Data Feedback API: Allows AI to receive real-time information on a character's posture, emotional state, and the location of environmental objects, ensuring more accurate AI decision-making and control.

What are the Benefits of AI Integration After Opening These Interfaces?
I've outlined a few points that I feel could help players, creators, and even the game ecosystem:

• Significantly Lowering the Barrier for Beginners: AI can act as a "smart assistant." We can describe our needs in natural language, for example, "make the character show a surprised expression and take a step back," and the AI can directly generate the corresponding complex motion, eliminating the need to struggle with complex operation tutorials.
• Improving Creation Efficiency: For experienced players and small studios, AI can automatically handle tedious, repetitive tasks like keyframe adjustments and IK/FK solving. Creators can focus more on story ideas and narrative expression.
• Enriching Game Interaction and Narrative: On one hand, NPCs will no longer be limited to preset dialogues; they can generate realistic body language based on context via the animation interface, leading to stronger interaction. On the other hand, ordinary players can easily create their own original stories, transforming the game from a platform for "playing content made by others" into one where "everyone creates together."

Finally, My Conclusion
Opening up AI-facing API interfaces is more than just lowering the operational barrier and enhancing user experience; I believe it is a crucial step for games to embrace new technologies and build a sustainable content ecosystem. AI-empowered creation is already a major trend, and I hope developers will pay attention to this point.
What do you all think of this idea? Have you played any games that support similar interfaces and allow AI-assisted creation? Or do you have other suggested needs or recommendations? Feel free to discuss together in the comments!

 

[techwoof]

Hello. I work full time as a developer and also study animation as a hobby, because i like it a lot, so i can understand your point and needs.

However, i feel like you are heavily misguided by what "AI" can do. First, suppose you had all the APIs in places, what kind of AI would be able to do such tasks? If you are talking about Language Models, that outputs text on inference, it simply do not have the capabilities you are looking for. Language Models don't have sense of physical space do to animations properly, nor the understanding of the feedback for such thing. It can, at maximum, drive/trigger predetermined parameters but all animations would need to be done by you first, the LLM would just trigger it's execution. For an example of this, see Voxta (https://hub.virtamate.com/resources/voxta-virt-a-mate-plugin.40039/) and VAMX (https://hub.virtamate.com/resources...custom-scene-personality-general-fixes.59564/), two big projects in VAM (VAM1 not VAM2 where this thread is created under) that does something similar, using language models to drive animations and speech. But as you can see, none of them are 'magical'.

Now, maybe you are not talking about LLMs but transformer-based models trained on mocap data for predictive animation. A software called Cascadeur uses this and makes it very easy to animate, assuming you know what you are doing. There is also this project who does something a little more 'independent' (https://github.com/Tencent-Hunyuan/HY-Motion-1.0). But none of them gives you a perfect job just by telling you what you want. At best, they can be used as facilitators.

Many people tried the very thing you are describing on your post, and there's a reason to why you haven't seen anything mainstream yet. It simply not possible, and how well it interacts in certain fields may give you the illusion current "AI" is way more capable than it actually is. Hooking it to an API is not all, there are many variables to account for, which makes just working with it hard enough, in the best possible scenario. There's no point in 'embracing new technologies' when they don't have the capabilities to do what you expect, and that's the case with "AI" right now for many tasks involving content production.

You can be sure somebody is spending a big money on it, and if you haven't seen something, it's simply because technology isn't there yet. For good creative content manual labor is required, otherwise it's just repetitive low-quality procedural content.

 

[hazmhox]

Asking for API/Technical implementations when saying you're a beginner and coming up with a machine assisted bullet point list is kind of very funny.