How computationally expensive would it be to run Neuro with 3D model through UE5?

[weary wren]

Video that sparked my question: https://www.youtube.com/watch?v=mR5ZtphHR_E

I just saw a video showcasing a pretty impressive "virtual avatar" in UE5. This made me think: What would it be like if Neuro had a model like this? However, I'm not too sure if it would realistically be feasible so I wanted to ask to programmers and those knowledgeable when it comes to VFX and stuff, would it even be possible for Neuro to use that kind of model without the PC imploding on itself? (CPU cost + GPU compute costs and stuff)

[jaunty jungle]

dunno but this is right up codemiko's alley

i think she was working on the infra of vtubing with unreal a few years ago (and now uses it)

[weary wren]

Yeah. I've seen some of her clips. But I guess, my main wonder was that since Neuro already uses so much power (on both the CPU and GPU) side, just to function, would a high end model even work?

I'll look more into it, I guess

[jaunty jungle]

i'm sure it's a fraction compared to all the local AI stuff that vedal is inferencing, given that videogames (which i guess use unreal) don't use that much VRAM

[weary wren]

I suppose that you are probably right. It would be interesting though if Vedal could try to pair Neuro to a 3D model like CodeMiko's though idk if the work would be worthwhile lol

[jaunty jungle]

"pair" in my mind implies having both the 2d and 3d models on screen - am i reading that right?