The meeting, already written up

Meeting notes

Playable below — no signup

Recording in, transcript out — with speakers identified, a summary you can send, and the action items separated from the conversation.

Open a sample meeting

How the privacy side works

Live demo

A finished meeting, exactly as the product shows it

Read the transcript, jump between chapters, rename a speaker, look at the decisions and the open actions. Same interface, sample meeting.

A made-up meeting between made-up people. The demo has no microphone access and no network access — nothing is recorded, uploaded or stored.

Open the real thing

What it produces

More than a wall of text

Speakers, not "Speaker 1"

Diarisation separates the voices; naming them once relabels the whole transcript. Optional voice profiles recognise the same person next time.

Actions and decisions, split out

What was agreed and what someone now owes, each with a timestamp back into the recording.

Summaries in your house style

Summary templates let a team fix the shape of the output — a retrospective, a client call and a stand-up should not read the same. Templates can be personal or shared across the organisation.

Seven engines, including one on your own GPU

Use a cloud transcription provider, or run WhisperX yourself so the audio never leaves the building. The choice is per organisation, and it is a configuration change, not a migration.

Recordings can arrive on their own

Nextcloud Talk · Google Meet

A Nextcloud Talk call can be recorded and ingested automatically. Google Meet recordings are imported once Google has finished producing them — Bee Flow does not send a bot into your call, it collects the recording afterwards. Everything else is an ordinary upload.

Because this is recorded speech

FAQ

The questions recordings raise

Where is the audio stored?

In your own object storage, alongside the rest of your data. You decide the retention; when audio is removed the note keeps its transcript and summary and simply says the recording is no longer available.

Do you need voice profiles to name speakers?

No. Diarisation separates speakers without them, and you can name each one by hand. Voice profiles only add recognition across meetings, and they are opt-in per person with their own consent step.

Can an admin access voice profiles?

No — by design rather than by policy. There is no read path for another person’s voice profile at any privilege level, and an organisation admin sees only how many people enrolled. Withdrawing consent deletes the profile. We never store the enrolment clip ourselves; be aware that enrolment sends it to pyannoteAI, which is why this feature is opt-in per person with its own consent step rather than on by default.

Which languages work?

The engines are multilingual; Dutch and English are what we use daily. Pick the engine that suits your language mix — that is a per-organisation setting.

Is this on the free tier?

No — meeting notes is an Enterprise module. The demo above is the whole interface, so you can judge it before talking to anyone.

Stop writing up your own meetings

Bring one recording and see what comes back.

Open the app Talk to us