Turn a recording into minutes with action items
An hour of audio goes in; a speaker-separated transcript, structured minutes and a who-does-what-by-when list come out.
What you end up with
- A verbatim transcript labelled by speaker
- Decisions and action items pulled out so nobody re-listens
- Audio stays local, if you use an on-device model
The sequence
1. Transcribe with speaker separation
Resolve who said what first. In a multi-person meeting, an unlabelled transcript forces you to re-listen, which defeats the purpose. For Chinese meetings, try a model tuned on Mandarin.
whisperOpen source★ 111.0KConverts speech to text and can also translate into English, supporting multiple languages and noisy environments.whisperXOpen source★ 24.8KAutomatically transcribes audio and video into subtitles with accurate word-level timestamps and tells you who is speaking.faster-whisperOpen source★ 26.0KA tool that accelerates the Whisper speech recognition model, letting you convert speech to text faster.2. Or use one tool that does transcription and minutes together
If you'd rather not assemble the pipeline, these turn audio straight into minutes at the cost of less control over the template. Fine for standups; for client calls you probably want to own the prompt.
meetilyOpen source★ 32.0KA locally running AI meeting assistant that provides real-time transcription, speaker identification, and automatic meeting minutes, with all data staying on your computer.buzzOpen source★ 22.1KOffline transcription of audio and video into text on your computer, with translation into other languages, without uploading to the cloud, protecting privacy.3. Force a fixed output structure
Give it a fixed skeleton: attendees / decisions / action items (owner + due date) / open questions. A fixed structure is what makes minutes scannable and pasteable into a task tracker.
Only need one piece of this?
You don't have to build the whole chain. Every tool above is listed on its own, with a summary, pros and cons, and a starting point.