Meeting notes
Bee Flow meeting notes turn a recording into a transcript per speaker, a summary, decisions and action items. Transcribe a Teams recording afterwards, or record a meeting at the table. Recordings from Google Meet and Nextcloud Talk come in too, and you choose the transcription engine: on your own server or a hosted one. From the Pro licence.
Plan a call See the plansThe real meeting notes screen with a sample meeting: the transcript per speaker, the summary and the action items.
Meeting notes demo
Sample data. No audio is uploaded or recorded, and nothing you do here is saved.
Recorded Teams and Google Meet meetings and Nextcloud Talk recordings come in without downloading and uploading them by hand. For Teams and Google Meet, no bot joins the call.
Upload an audio or video file, or record a meeting at the table in the browser.
Every meeting gets a summary, the decisions that were taken and the action items that were agreed, next to the full transcript.
The transcript is split per speaker. Names come from introductions, from people being addressed and from the attendee list, and you can rename a speaker by hand.
With the pyannoteAI engine, a colleague can record a voice profile once. Bee Flow then puts their name on their own turns. A speaker is only named when the match is confident.
Voxtral (Mistral), WhisperX (self-hosted), Scaleway Whisper, Azure AI Speech, Azure Whisper batch, pyannoteAI and a built-in Whisper on CPU. The built-in Whisper is a small model with a default cap of about ten minutes per recording. An administrator can limit which engines may be used.
Where the audio goes
The engine you choose decides whether audio leaves your server. Meeting notes need the Pro licence.
WhisperX runs on your own server and is the engine for a firm that wants every meeting on its own hardware. It needs a capable server. The built-in Whisper also runs on your own CPU, as a small model with a default cap of about ten minutes per recording.
Voxtral (Mistral), Scaleway Whisper, Azure AI Speech, Azure Whisper batch and pyannoteAI are hosted services. Audio you send to them leaves your server. Voice profiles work only with pyannoteAI.
The summary, the decisions and the action items are written by the AI model your organisation set up: a hosted one, or a model on your own hardware.
Import it from Teams, Google Meet or Nextcloud Talk, upload a file, or record in the room.
The engine your organisation set up writes the transcript and splits it per speaker.
Read the summary, check the decisions and pick up your action items. The full transcript stays alongside.
Forward this page. The facts they will ask for are written down, including what is missing.
Prompts, files and answers: where they go, who can reach them, which model providers are involved and what is logged.
Sign-in, roles, retention and encryption, and what is not covered.
One command with Docker on amd64, or Bee Flow Cloud in the EU.
Yes. A meeting that was recorded in Teams comes in afterwards through the organiser's own Microsoft connection, and no bot has to join the call. Bee Flow reads the recording, or the transcript Teams made. An administrator can also set this up for the whole organisation. That takes consent from an Entra administrator and recording allowed in the Teams policies.
Yes. Record the meeting in the browser on a laptop on the table, with pause and resume, or upload an audio or video file afterwards. The transcript is split per speaker. Tell the people in the room that the meeting is being recorded.
By the engine your organisation chooses. WhisperX runs on your own server. The built-in Whisper runs on CPU on your own hardware. It is a small model with a default cap of about ten minutes per recording, with a cloud engine as fallback for longer recordings. Voxtral, Scaleway Whisper, Azure AI Speech, Azure Whisper batch and pyannoteAI are hosted services, so audio sent to them leaves your server.
Microsoft Teams, Google Meet and Nextcloud Talk. Teams and Google Meet meetings come in through the organiser's own Microsoft or Google connection, and an admin can also set up Teams for the whole organisation. A Google Meet meeting can only be imported when it was recorded in Meet. There is no Teams chat integration.
A voice profile lets Bee Flow recognise a colleague by voice. It works only with the pyannoteAI engine, a hosted service that receives the audio. Only you can record your own, with separate consent, and withdrawing that consent deletes it. The recording itself is not kept: only an encrypted profile is stored. People without a profile are still split per speaker, as before.
Meeting notes are part of Pro, and Enterprise includes them too. Chat, assistants, knowledge bases and the automation builder are already in Free.
No. Zero-knowledge encryption covers private chats. In new organisations the transcript, the summary, the speakers and the attendees are encrypted with the organisation's managed key, which the server holds, because transcription and search have to read them. Three fields fall outside that encryption: the action items, the decisions and the open questions. Self-host Bee Flow if that server has to be yours.
The NBA and NOREA AI guide, condition by condition, next to the setting in Bee Flow.
What happens between the recording and the action list, step by step.
Which engine keeps the audio inside, and what it needs.
Talk recordings, files and calendars, under each user's own account.
Tell us which meeting tools you use and where your audio may be processed. 20 minutes. An honest go or no-go, and what a fixed price would look like.
Plan a call Try the live demo