whisper-stt-plus-timestamps.md

Whisper STT + timestamps

https://github.com/dmarx/video-killed-the-radio-star

---

https://colab.research.google.com/github/openai/whisper/blob/master/notebooks/LibriSpeech.ipynb#scrollTo=T5Gs4qD52D46

use grad cam on output tokens to infer per-token timestamps

see also: https://github.com/dmarx/bench-warmers/blob/main/subtitles-to-storyboard.md