| From TFA: > Can’t you just generate an image of a face, animate it, and add a voiceover? Not quite. The challenge isn’t just nailing the lip sync — it’s making facial expressions and body language move in tandem. It would be weird if your mouth opened in surprise, but your cheeks and chin didn’t budge! Starting here: - "generate an image of a face, animate it, and add a voiceover?" Tried this at [0]. Here's an example visual output: - https://visualmic.com/example-animation-4.gif Judge for yourself. You can see the mouth and eyebrows move in response to voice volume, and the eyes shift and blink according to settings. But no cheek movement, no head tilt, and no face shape change. I think TFA is sort of right. I'm not sure that face cap and AI are 100% needed, and most of the tools for making great VR models seem either pretty complicated or sort of privacy invading. But, better translating voice input into face changes does seem sort of needed. There's "virtual Youtuber" (vtuber) software too, but that too seems some combination of complicated/clunky, resource intensive, and/or in signup-required land. [EDIT: Surely, there is a good front end at OpenLive3D [1], but making the .VRM model for it, e.g., with VRoid Studio [2] is where things seem start to get a little more time/energy-intensive.] I'm not against pseudonymous avatars, but is there a third path? It should be easy and open, no? Gonna have to trawl through the suggestions of a16z on this one. [0] https://visualmic.com
[1] https://openlive3d.com
[2] https://vroid.com/en/studio
|
