FFMPEG preprocessing for Wav2Vec2 - lmmx/devnotes GitHub Wiki

To convert an MP4 to WAV (resampled at 16kHz):

ffmpeg -i output.mp4 -ar 16000 -ac 2 -f wav output.wav

To split this into 5 minute chunks:

ffmpeg -i output.wav -f segment -segment_time 300 -c copy segmented/output%09d.wav

(Note the filenames will have 9 digits for the number here, probably excessive)

E.g.

output000000000.wav

If you choose a time period too short, you'll see the error that

RuntimeError: Calculated padded input size per channel: (2). Kernel size: (10). Kernel size can't be greater than actual input size

• See also: FFMPEG in Python