Meeting Notes Summarizer
Transcripts to action items in 30 seconds
Feed raw meeting transcripts or recordings through a speech-to-text layer, then extract a structured summary with key decisions, action items with owners, open questions, and next steps — delivered to participants before they leave the room.
GPT-6 Luna
for 500K tokens a month · 85% input / 15% output
Why this model
For this workload — 500K tokens a month, 85% of them input — it comes to about $0.08 a month, the lowest of the models suggested here. Its 1M-token context window holds long inputs in a single call, which matters for this kind of work.
Alternative models
Implementation tips
- 1
Structure output with explicit owner fields for every action item and a due-date field — vague action items like 'discuss this later' contribute nothing; the model should be instructed to flag these and ask for clarification if the transcript doesn't specify.
- 2
Process in two passes for longer meetings: first generate a bullet timeline of key moments with timestamps, then extract action items from the timeline — this handles 2-hour meetings that would otherwise exceed context limits cheaply.
- 3
Integrate with your project management tool via API: auto-create tasks from extracted action items with the owner and meeting context pre-populated, closing the loop from meeting to task in seconds.