Your notetaker,
minus the bot.
No awkward bot joins your meeting. The Notetaker captures the audio right in your browser, transcribes it live, and writes a structured summary the moment you stop.
Three steps,
nothing recorded.
Audio is captured in your browser, streamed to a transcription engine, and turned into a structured summary. The waveform passes through — it's never saved. What you keep is the transcript and the notes the AI writes from it.
Mic only, or mic plus the meeting tab's audio. Linear16 PCM, mono, 48kHz via an AudioWorklet — no plugin, no download.
The stream is transcribed as it arrives — final text locks in while the next words show greyed and interim, then settle.
When you stop, the transcript and your notes go to the AI, which returns eight structured fields — ready to edit and send.
One toggle.
Mic, or mic + tab.
For an in-person chat, the mic is all you need. For a video call, capture the meeting tab's audio too — the browser hands over the sound, and we throw the video away. If the tab picker is cancelled, we quietly fall back to mic-only so you're never left without a transcript.
Watch it
keep up.
Final words lock in; the next phrase shows greyed while it settles. Jot free-form notes alongside the transcript, and keep your prep brief one click away. Everything auto-saves — notes every two seconds, the transcript every fifteen.
40 seats. Wants pricing + a technical demo with IT lead (J. Park).
Flag: rollout time is the real objection.
Eight fields, written for you.
Edit it,
then send it.
Everything is yours to change — transcript, notes, summary and email all open in a markdown editor. Turn action items into tasks, send the follow-up, and find any session later by what was actually said.
Transcript, notes, summary and email all open in a markdown editor — fix a name, tighten a line, restructure a list.
Each action item becomes a real task with one click — owners and due dates carry across.
Sends via Outlook, falling back to Gmail. Your HTML signature is appended at send time.
Search every session by title, summary or transcript, and filter by status to find anything fast.
If a listening session crashes mid-call, the orphaned row is easy to finalize or discard — nothing left hanging.
The raw audio was never stored. What you edit, send and search is text you can see — and delete.
Built to be safe, and to keep up.
The boring engineering that makes the magic dependable — short-lived keys, auto-reconnect, and a hard rule that audio never lands on disk.
The waveform is transcribed in real time and discarded. Only the transcript persists.
Browser sessions mint a 60-second key (usage:write scope). The master key never touches the client.
Drops retry with 1 / 2 / 4 / 8s backoff, plus 8-second keep-alive pings to hold the stream.
Transcript capped at 100k chars (keeps the tail), notes at 8k, with a 60s summary function limit.
Just talk.
We'll take it down.
Turn on the Notetaker for your next call — no bot, no setup, no recording left behind. Free for fourteen days, included on every plan.