No output generated when using options --vad and --demucs

[dgoryeo]

Hi,

Running CLI command with either of options --VAD or/and --demucs results in no output. However without those options everything works. What am doing wrong?

Thanks

[Hemangpandey]

@dgoryeo may I know on which point you are using these parameters?

[dgoryeo]

Hi @Hemangpandey , this is the command line I tried --in Jupyter:

`
!stable-ts S01E04d-16khz.mp3 --language Japanese --model large --task transcribe --demucs True --vad True --debug -o S01E04d-16khz.srt

`

Thanks

[Hemangpandey]

@jianfch may I know how to use these paraments in python script and where?

[chislon]

Hi @Hemangpandey , this is the command line I tried --in Jupyter:

` !stable-ts S01E04d-16khz.mp3 --language Japanese --model large --task transcribe --demucs True --vad True --debug -o S01E04d-16khz.srt

`

Thanks

Tried this command locally on a Windows PC and it works.

[dgoryeo]

Thanks for testing @chislon , I believe that means that there might be a build/dependency issue in my environment. I'll try a re-install to check.

[jianfch]

@Hemangpandey It's part of transcribe(). You can check docstring to see what they do.

[jianfch]

The VAD might have marked the entire audio track as silent or Demucs might have removed all the sound from audio making it silent.

[Hemangpandey]

one more question about this ,suppose in the input there is mix of both music and normal conversation ,so still we should add both the parameters? @jianfch

[jianfch]

@Hemangpandey it should work fine too. The goal of both is to isolate the voice from other sounds before feeding into Whisper.