Open Cyanwise opened 2 years ago
I was able to find a hacky fix to get it working. CTRL+F and look for the following code in the google collab:
if selected_video.startswith('user/') or selected_video == 'demo/videos/0.mp4':
imageio.mimsave('temp.mp4', [img_as_ubyte(frame) for frame in predictions], fps=fps)
FFmpeg(inputs={'temp.mp4': None, selected_video: None}, outputs={'output.mp4': '-c copy -y'}).run()
else:
imageio.mimsave('output.mp4', [img_as_ubyte(frame) for frame in predictions], fps=fps)
And change it to just this:
imageio.mimsave('output.mp4', [img_as_ubyte(frame) for frame in predictions], fps=fps)
Hello, I am the author of demo.ipynb
. The line with FFmpeg
is supposed to copy the audio from the source video to the output. Did it do anything unexpected? @AbstractVoid
Whenever I try to make a deepfake, it gives me the video I use to animate the face instead of the result it supposed to give me. It never happened to me in the old demo.