decode_audio error about faster-whisper

[guillaumekln]

We use PyAV for the audio decoding which is based on FFmpeg, so it virtually supports all audio codecs.

I'm not sure this error is related to the file duration. Is it possible for you to share this input file (or any other file that triggers this error)?

There is a similar error reported here PyAV-Org/PyAV#1048 which seems to happen when the audio sampling rate is changing.

[Latrolage]

Is it possible for you to share this input file

I downloaded it from youtube yt-dlp -x -f bestaudio "https://youtu.be/P4xuSCEjD5Q" -o "audio.%(ext)s"

just running ffmpeg -i audio.opus audio.mp3 to re-encode it to lame stops the error from happening. using --format-sort acodec:m4a to download the m4a from youtube also works.

I'll close the issue since it seems that it's an issue in PyAV

[guillaumekln]

It seems we are hitting a size limit when pushing audio frames in av.audio.fifo.AudioFifo.

I updated the function decode_audio to avoid the error on long audio files: 2646906