There's a treasure trove of content in PyTorch dev podcasts, that I appreciate.
I find text faster to follow than than audio, so I'm just transcribing it using whisper.cpp
python3 download.py
bash to-wav.sh
bash transcribe.sh
I mostly center the subtiles and leave it to play using mpv
:
$ mpv --sub-file=XLA.wav.srt XLA.mp3 --force-window --sub-pos=50