search

vskadandale / vocalist

Official repository for the paper VocaLiST: An Audio-Visual Synchronisation Model for Lips and Voices
Other
61 stars 7 forks source link

Can the test result be a video? #4

Closed c1a1o1 closed 2 years ago

c1a1o1 commented 2 years ago

Can the test result be a video?

vskadandale commented 2 years ago

Your question doesn't make sense. The result is accuracy. In case you want to know if the model can be tested on a video, you need to process the video in the same way you process LRS2 dataset as done in Wav2Lip paper. The code to test directly on a video is not available as such in this repository. However, it is very straightforward to do that if you follow the preprocessing steps from Wav2Lip paper.