had a crack at some auto-phoneme-generated lipsync this arvo. I think the phonemes themselves are alright and the jank is mainly the shapes and blending, but also this is using Rhubarb which is meant for 2d stuff and only does 9 phonemes

the goal would be something comparable to HL2, whose basically-automatic phoneme extraction is extremely good and detailed. You might as well just mocap your whole face with a phone these days but not practical for every line probably, nice to have a passable lipsync for any wav

https://youtube.com/shorts/nSHnQ58wkyc?si=r5tTxhLCyanD0DUL