search

primepake / wav2lip_288x288

MIT License
555 stars 138 forks source link

custom dataset #24

Closed EhsanRusta closed 2 years ago

EhsanRusta commented 2 years ago

Hello @primepake, pls help me with dataset correction, on my dataset FPS=30 and need to changed it to 25, up to now I did the face detection and the preprocessed data is frames and audio, what should I do to change FPS? thank you

ghost commented 2 years ago

I need to convert it to 25 fps before detecting faces

EhsanRusta commented 2 years ago

and what about if I want proceed with 30 fps?

ghost commented 2 years ago

this fps will effect your model, you need to scale up model and change hyper paramertes: syncnet_T, syncnet_mel_step_size you need to change the hyper-parameters correctly with 0.2 seconds

EhsanRusta commented 2 years ago

also my data sample_rate of audio is 44100 and not 16000, should I change any hyper parameters?

ghost commented 2 years ago

read https://github.com/primepake/wav2lip_288x288/issues/21#issuecomment-1172847918

EhsanRusta commented 2 years ago

thanks i will change both fps and sample rate and also Im facing this error for syncnet, did you experienced this? error