Hello creators and consumers, I've been thinking about artificial intelligence for VAM and am wondering if anyone has made any progress in a few key areas. Most of the discussion around AI on here seems to be geared toward using large language models to simulate conversation or to upscale VAM images to be more realistic or filtered, but I'm wondering if anyone has started using some of the art or coding focused uses of artificial intelligence to make things in VAM? Specifically, I'm wondering about:
- Textures: It seems like imagery focused AI could easily be used to make crisper textures for VAM characters and assets. I found an interesting tutorial on this here and wonder if anyone has used a similar approach? . See also here for some Unity specific texture making. https://www.cgchannel.com/2023/07/check-out-three-new-generative-ai-add-ons-for-unity/
- Plugins: As we know, ChatGPT is capable of writing code once you get it used to the parameters you're working with. Because VAM is such a specific use case, I'm not sure exactly how it could be useful for plugin creation, though I'm wondering if there are ways it can assist with tweaking existing plugins or if it can make tweaks using BepInEx that mod the existing Unity architecture? I should note there are some drawbacks to using AI for coding as chronicled here: https://www.zdnet.com/article/how-to-use-chatgpt-to-write-code/
- Morphs: Since morphs rely on simple numerical value tweaks, I wonder if it's possible to use AI to add randomness to morph value changes. Here specifically I'm thinking of having a figure turn from young to old spontaneously in a scene or have their features shift like a shapeshifter. Think Mystique from the X-Men or the Skrulls. Rather than a full appearance preset switch, this would be morph movement in real time while you were looking at the character. This would also allow for random size and shape changes dependent on either a trigger or a time interval. I know some version of this has been done in the past to switch a character progressively from male to female and vice versa, albeit not with AI.
- Skydomes: We've already seen some people using AI generated art for Skydomes and of all the uses, I think this one has the most potential, particularly if you could somehow have a voice activated AI that makes a skydome or photosphere for you on demand. Some more on this here:
- CUAs: It seems natural language processing is already being used to create 3d environments, but I'm not sure how detailed they can be or whether it's not just easier to continue the practice of importing free sketchfab templates into Unity and on to VAM. I'm curious whether anyone has used text based prompts in something like Spline or something else to build environments or at minimum textures for CUAs?
- Animations: There are a lot of great sex animations on here, but not a ton of other movements like walking, jumping, running, or falling down. These exist, but are more complicated than they look to replicate. What I wonder is whether it would be possible to use machine learning to do a batch conversion of existing BVH files to be DAZ G2 compatible or make BVH files more physics responsive? This was discussed somewhat here: https://hub.virtamate.com/threads/i-found-ai-machine-learning-to-convert-video-to-mocap.7609/
- Particles: Here I'm thinking how cool it would be if there were some way to show the effects of particles on a object or character so if I were using lightning or a flamethrower on a character, VAM characters would be responsive to the particles or potentially they could trigger an adjustment of morphs and textures to show that they interacted with the particle. Here I'm thinking of something like this: https://medium.com/nerd-for-tech/gaming-ai-unity-with-deep-reinforcement-learning-e8be9d14f248
