| Granola are cracking down on their auth and I can no longer pull my meeting notes out and put them in Obsidian (my plugin is gonna have to be deprecated: https://github.com/tomelliot/obsidian-granola-sync/). That means I'm gonna need a different meeting transcription/summary pipeline. There's a cambrian explosion of meeting transcription tools (https://hn.algolia.com/?q=Show+HN%3A+transcription). It's awesome. I've tried some, but haven't found one that works how I want. Can you help me find one that suits my use case? - daemon runs in the background - uses local models for transcription, diarization/speaker identification - primary interface is CLI/no UI. It should just happen in the background. - GUI is secondary. Maybe a button in the menu bar for starting/stopping a meeting optional/secondary - pulls in calendar info for titling and attendees - cleans the transcript - generates summary |
Speaker identification is a huge trick to do on device. I've masked by sentence and capture embeddings within that mask to get things more accurate.
If you try it that would be super cool and it's basically free rn while I figure out how to market.
I'm testing obsidian integration rn. It's currently available through mcp but that kind of defeats the offline transcription thing