audion sync and lips sync
#2
by ttamementomori - opened
Thanks the lora is great for the moves !! but I noticed it doen't seams to take the original audio in the process and it generate a new one breaking lipsync with with original audio is there a way to fix that ?
ttamementomori changed discussion status to closed
I didn’t test it, but it will probably work:
Instead of generating the audio, “freeze” it to be your input audio.
Specifically: the audio is part of the generation, but it’s not being regenerated- ans the model will use it in its context
thats the weird thing i encode original audio instead of the blank default one but it generates a fresh audio. anyway the moves are working great !!!
edit : it works using this node for the audio latent : 🅛🅣🅧 Set Audio Ref Tokens