how to extract whisper encodings

[nkumar248]

Hey, I wanted to ask what the best way to extract whisper encodings is. I have tried using a fork by @isaacOnline but it is giving me confusing results (have previously posted a question on this) so I'll try posting an open question.

[montesclarosglennbenedict]

You could try using the SpeechBrain library on it. I forget the implementation but it offers an alternative to the Isaac's fork.

In general, you want to process it based on [OpenAI's Cookbook] (https://cookbook.openai.com/examples/whisper_processing_guide), where you

1. Trim the audio to the relevant parts
2. Add relevant terminology to the encodings
3. Fix any unicode issues with the encodings