Know who said what, and what happens next.
MeetingScribe turns a meeting recording into a transcript with speaker labels, a short summary, and a list of action items. It handles English mixed with Swahili.
Desktop app for Windows now. macOS is next.
Tutaanza saa tatu tomorrow, but we need to confirm the venue first.
I’ll call the hall today. If it’s booked, we move to the office.
Good. Please send the agenda to the whole team by Friday.
Summary
The team will start at 9 a.m. tomorrow once the venue is confirmed. The office is the fallback.- Call the hall to confirm the booking
- Send the agenda to the team by Friday
From recording to action items in one pass
You don’t have to tidy anything. Summaries and action items appear automatically after every transcription.
Record or import
Record from your microphone, or import MP3, WAV, OGG, FLAC, AAC or Opus files.
Transcribe
Speech recognition runs on our own servers on AWS. Audio is not sent to a third-party transcription API.
Label speakers
Each paragraph is tagged Speaker 1, Speaker 2 and so on. You can tell it how many people were in the room.
Summarize
A short summary and the action items are written for you, then copied or exported in one click.
Built for meetings the way they really happen
Speaker labels
Multi-person meetings come back as readable paragraphs, grouped by who was talking and broken at natural pauses.
English and Swahili together
Many teams switch languages mid-sentence. MeetingScribe uses a multilingual model and a vocabulary hint for local place names, so it doesn’t force everything into one language.
Action items you can trust
The summary only lists tasks that were actually said. If a meeting has no tasks, the list stays empty rather than inventing some.
You control the data
Recordings and transcripts live in a private storage bucket in your cloud account. Access is limited by a dedicated role scoped to that bucket.
Built on AWS
MeetingScribe runs end to end on AWS. Credits would let us move from a single test server to a service that real teams can depend on.
| Amazon EC2 | Hosts the transcription and speaker-labeling service (Whisper and pyannote behind a FastAPI endpoint). |
|---|---|
| Amazon S3 | Private storage for uploaded audio and finished transcripts. |
| Amazon Bedrock | Claude Haiku 4.5 writes the summary and extracts action items using structured tool calls. |
| AWS IAM | A dedicated role gives the server access to one bucket and nothing else. |
| Region | eu-north-1 (Stockholm), with every resource tagged to the MeetingScribe project. |
What’s working and what’s next
We’re building one feature at a time and testing each with real recordings.
Transcription
Record or import audio, speaker labels, automatic summaries and action items in a Windows desktop app.
Meeting recording
Capture online meetings directly, so you don’t need a separate recorder.
Shared to-do list
Action items flow into a task list that syncs across your team.
Try MeetingScribe on your next meeting
We’re onboarding a small group of early users. Tell us how your team meets and we’ll get you set up.