> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kombify.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Record and review a meeting

> Dual-channel meeting capture, notes, screenshots, and cited write-ups

Meeting records a call as two independent channels — your microphone and the
system loopback — transcribes both into one session, and writes the meeting up
from the transcript, your own notes, and a template. It is a separate recording
service, not a fourth mode: it shares only the speech-to-text transcriber with
Dictation.

<Info>
  Meeting is a Windows desktop capability. System-audio capture is not ported to
  the macOS beta.
</Info>

## What a meeting produces

* A transcript with both channels, labelled by capture channel. Channel labels
  say which side of the call a line came from; they are not identification of
  every remote speaker.
* Your own notes, reproduced verbatim in the write-up.
* A durable **Meeting Review**: a five-sentence executive brief, action items,
  and a body that cites the transcript it came from.
* Reviews in English, German, Spanish, Chinese, Hindi, and Arabic.

Audio is never persisted. The transcript, notes, and review are what remain,
and [Retention Scope](/speechkit/modes#retention-scope) decides how long.

## Start and finish a recording

1. Start a meeting from the desktop client, or let call detection offer one
   when you accept a call. Detection is a suggestion you accept, not an
   automatic recording.
2. A compact transparent overlay shows capture health and controls during the
   call, and can unfold for notes.
3. Rolling summaries run in batches while capture continues, so a long call
   does not wait until the end to be digested.
4. SpeechKit stops shortly after the calling application releases the
   microphone, and notifies you when the review is ready or why it failed.
   A recording you started by hand is never stopped automatically.

Exactly one capture holds the microphone at a time. Starting Dictation, Assist,
or Voice Agent while a meeting records is refused rather than silently taking
the device.

## Screen snapshots

Press `Ctrl+Alt+S` during a meeting to capture a chosen window or the monitor
under the cursor. Snapshots appear as markers in the transcript timeline, are
cited in the generated review, and briefly preview in the meeting pill so you
can confirm the right window was captured. No image leaves the device.

## Find and export

The meeting list groups sessions by when they happened. Search covers titles,
transcripts, notes, and write-ups. Notes render as formatted Markdown, and the
complete write-up offers **Copy Markdown** and **Save .md**.

## Review generation

The local path needs no cloud credentials: the bundled local model generates
the review on the device. GitHub Copilot is an optional review-generation
provider behind a separate, revocable transcript-processing grant.

Export connectors and cloud review providers are unavailable under the
**local network only** and **this device only**
[Network Scopes](/speechkit/modes#network-scope). A fresh install configures no
connector, so nothing leaves the machine until you choose it.

<CardGroup cols={2}>
  <Card title="Install on Windows" icon="windows" href="/speechkit/install-windows">
    Meeting ships with the Windows desktop client.
  </Card>

  <Card title="Modes and data paths" icon="waveform-lines" href="/speechkit/modes">
    See how provider choice and the two privacy scopes work.
  </Card>
</CardGroup>
