search

SamKhoze / ComfyUI-DeepFuze

DeepFuze is a state-of-the-art deep learning tool that seamlessly integrates with ComfyUI to revolutionize facial transformations, lipsyncing, Face Swapping, Lipsync Translation, video generation, and voice cloning.
https://cogidigm.com/
Other
290 stars 32 forks source link

Can not reptoduce the face and eye movement. #11

Open OrvilleQ opened 3 months ago

OrvilleQ commented 3 months ago

Hello.

I am trying to reproduce the character's face as well as eye movements in the demo video you provided. Since the female hair in that video is floating all the time, I assumed that it was generated from a single still image. So I generated a similar image and expanded it to 240 batches and generated it based on the image examples in the repository. However, the generated video only has the mouth movement, not the character's face and eyes following the speech as in the demo video.

How can I achieve the effect of the demo video?

Thank you for your help.

SamKhoze commented 3 months ago

@OrvilleQ we are adding a new node that is based on SadTalker and you can generate video from an image, however, it only works with close-up faces, the node will be added shortly