MuseTalk
MuseTalk copied to clipboard
Model training
I trained 100,000 steps for 24 hours on a 5-minute conversation, but the result does not differ in any way from the original model, what could have gone wrong? the loss decreased as the training progressed.
And one more question, I downloaded the HDTF dataset, but the video is there without sound, is it possible to train a model on video without sound?
And one more question, I downloaded the HDTF dataset, but the video is there without sound, is it possible to train a model on video without sound?
I don't think it is possible.