How Note Genie Captures Meetings

The technical architecture behind Note Genie's meeting capture stack — and why it's different from every other AI notetaker.

Last updated: April 2026

The short version

Most AI notetakers use one method to capture every meeting — usually a participant bot that joins as an attendee, records the audio in a fake browser tab, and runs its own transcription. That works, but it has obvious downsides: a bot in your participant list, a noticeable join delay, and unreliable speaker labels.

Note Genie uses six different capture methods. A central Capture Coordinator picks the best one for each meeting per user, automatically. When a native API integration is available we use that — no bot, no DOM scraping, native speaker labels, and (for Microsoft Teams and Google Meet) the meeting transcript is generated by Microsoft or Google for free, which means we never run our own AI transcription on those meetings. The result: lower cost, higher quality, and total invisibility on the meetings where it matters.

The capture tiers

Each tier is independently developed and tested. The coordinator routes every meeting through exactly one tier — never two, so you never see duplicate recordings in your library.

1. Privacy mode

When you turn on Privacy Mode in your Note Genie settings, the coordinator refuses to dispatch any cloud capture — no bot, no native API, nothing. Audio stays on your device until you choose to upload it via the desktop app or the Chrome extension. Read the full guide at /docs/privacy-mode.

2. Zoom Realtime Media Streams (Zoom RTMS)

For Zoom meetings hosted by users who have connected the Note Genie RTMS Marketplace app. Zoom streams the meeting audio directly to Note Genie via WebSocket — no bot joins the meeting, no participant slot is taken, no DOM scraping. We receive per-participant audio with real speaker labels from Zoom's metadata. Read /docs/zoom-rtms.

3. Microsoft Teams (native Graph integration)

For Microsoft Teams meetings hosted in any Microsoft 365 tenant whose admin has installed the Note Genie Teams app. We subscribe to Microsoft Graph change notifications on the tenant's recordings and transcripts, and Microsoft fires us a notification the moment the meeting ends and the recording is ready. We download the recording + Microsoft's structured transcript directly via the Graph API. Microsoft generates the transcript for free, so we never run our own AI transcription on Teams meetings — that's a real cost saving we pass on to you. Read /docs/teams.

4. Google Meet (native Workspace Events integration)

For Google Meet meetings hosted by users who have connected their Google account. We subscribe to Google Workspace Events on the user's Meet recordings and transcripts, and Google fires us push notifications via Cloud Pub/Sub. We download the recording from Drive and Google's structured transcript entries from the Conference Records API. Like with Teams, Google generates the transcript for free, so we skip our own AI transcription on Google Meet meetings. Read /docs/google-meet.

5. Note Genie cloud notetaker (the fallback bot)

For everything the native integrations don't cover: Webex, BlueJeans, Whereby, Zoom users without RTMS, free Gmail accounts (Google Meet API access requires a Workspace plan), Teams meetings outside a connected tenant, and the long tail of niche platforms. Our notetaker joins the meeting as a visible participant, captures the audio, and uploads it to our processing pipeline. This is the only tier with a visible bot — and it's the only tier we use when none of the native paths are available.

6. Local recording

For users who want zero-cloud capture by choice or by policy. The Note Genie desktop app captures system audio at the OS level (ScreenCaptureKit on macOS, WASAPI loopback on Windows, PipeWire loopback on Linux) — anything playing on your speakers can be recorded locally to a WAV file. The Note Genie Chrome extension captures browser tab audio via the Chrome tabCapture API. Both are local-first: audio never leaves your device until you choose to upload it.

How the coordinator picks

Two values drive the routing decision for every meeting:

The rule

Pick the tier with the highest quality score. When two tiers tie on quality, pick the one with the lowest cost. Privacy mode is an unconditional override that bypasses everything.

In practice: Microsoft Teams and Google Meet meetings always route to the native integration (when available) because it's both highest quality AND lowest cost. Zoom meetings route to RTMS (when available) for the highest quality. Everything else falls through to the cloud notetaker bot. Privacy mode disables all of it and forces local-only recording.

Why this matters

Lower cost passed on to you

Most AI notetakers run their own transcription on every meeting and bake that cost into your subscription. We don't need to. On Teams and Google Meet meetings — which are roughly half of all enterprise meetings — Microsoft and Google generate the transcript for us at zero cost. We pass that saving on to you in the form of lower per-user pricing than competitors who pay for transcription on 100% of meetings.

No bot when a native path is available

Notetakers that always send a bot create real friction: it takes a participant slot, sometimes triggers a join sound, and sometimes makes meeting hosts uncomfortable. With Note Genie, when you connect a native integration, no bot joins your meetings on that platform — ever. You get the transcript and AI summary in your library after the meeting ends, with no in-meeting indicator that recording is happening.

Higher transcript quality

Native integrations give us per-participant audio streams with real speaker labels from the platform itself. That means "Alice said X, Bob said Y" — not "Speaker 1 said X, Speaker 2 said Y". The cloud notetaker bot only gets a single mixed audio track and falls back to AI diarization, which is ~80% accurate at best. Native is ~99%.

Reliability that doesn't break when platforms update

Bot-based notetakers join meetings via headless Chrome instances driven by Playwright, which means they have to scrape the DOM to find the "Join" button, the "Mute" button, etc. When Zoom or Teams or Google Meet update their UI (which they do every few weeks), the scraping breaks and the bot fails to join. Our native integrations don't scrape anything — they use official APIs that Microsoft, Google, and Zoom commit to maintaining.

How privacy mode interacts with the tier model

Privacy mode is the only tier that overrides everything else. When it's on, the coordinator skips even the passive native integrations — because even those send your audio to Microsoft, Google, or Zoom for processing. With privacy mode on, your audio never leaves your device until you choose to upload a recording manually. Read the full privacy mode guide at /docs/privacy-mode.

The technical spec for IT admins

If you're an IT admin evaluating Note Genie for your organization, here's the short technical spec you can put in front of security review:

Questions

Email support@notegenie.org or talk to your account manager. We're happy to walk through how the architecture applies to your specific compliance requirements.