[WIP] feat: add channel check on the audio

[MENGZHEGENG]

PR Goal?

Add check on the number of channels in the provided audio, remind the user when any audio has more than 2 channels (which can be problematic) and exit the program. Otherwise we will keep the only channel or average if there are two channels.

Fixes?

Fixs #351

Feedback sought?

Priority?

Tests added?

TODO

How to test?

Confidence?

Version change?

Related PRs?

[semanticdiff-com]

Review changes with  

Changed Files

File	Status
everyvoice/wizard/basic.py	0% smaller

[joanise]

Question about this test, counting channels: according to https://docs.python.org/3/library/wave.html, wave and wave.open() only support "uncompressed PCM encoded wave files".

@roedoejet Is EveryVoice already only compatible with such wav files, or could using wave.open() break some existing audio format flexibility?

I know that wav2vec2aligner can take .m4a audio files and align them, I've tried that before with success, but I don't know if they're supported anywhere else.

[roedoejet]

Question about this test, counting channels: according to https://docs.python.org/3/library/wave.html, wave and wave.open() only support "uncompressed PCM encoded wave files".

@roedoejet Is EveryVoice already only compatible with such wav files, or could using wave.open() break some existing audio format flexibility?

I'm not 100% sure, but we use torchaudio.load elsewhere in EveryVoice to load and save audio, so I think we should use that here too