I've added audio as another interface for regular and useful writing—across daily notes, Git comments, etc. I thought I'd quickly package it up and make it available¹. I shared the snippets originally with my colleagues. It uses an MLX version of a Whisper model.

It's helpful because at the moment it takes 3 sources:

1. Record from a microphone
2. Download a YouTube video
3. Provide a path to a local audio file

There's main commands, depending on what I'm doing. Sometimes I want a transcription verbatim. Other times, I want a nice summary. I've added a profile (or, system prompt), to style them specifically depending on my audience. For example, for pull requests.

I've already used this a bunch of times! Try it out yourself¹.