Recording a Meeting

The Meetings tab is designed for longer recordings where you want a full transcript, speaker attribution, and optionally an AI-generated summary. It is separate from the dictation flow — nothing is auto-pasted anywhere.

Starting a recording

1. Open Murmur and go to the Meetings tab.
2. Optionally enter a meeting title. If you leave it blank, Murmur will generate a title from the transcript after the recording ends.
3. Select your transcription engine from the dropdown. The currently active model is shown by name.
4. Click Start Recording.

A History entry for the meeting is created the moment you start, so the meeting exists — and is recoverable — from the very first second. (New in 1.6)

During a recording

While a meeting is being recorded, the Meetings tab shows a pulsing Recording indicator with:

• Duration timer — advances every second (HH:MM:SS), even if audio momentarily stalls. (New in 1.6 — it no longer freezes when input goes quiet.)
• Audio level meter — a live waveform of microphone input.
• Model name — a reminder of which transcription engine is active.

The menu bar also shows a red dot and running timer, visible across all spaces and apps.

Microphone changes are handled for you

(New in 1.6) Meetings use Murmur's shared, hardened audio engine, so a microphone change mid-meeting — AirPods connecting, a cable nudged loose, an interruption — no longer ends the recording. Capture re-acquires the new device (following your Microphone Priority ladder) and continues. During the brief gap the bar shows "Reacquiring audio source."

Live warnings

(New in 1.6) If something needs your attention, an orange warning appears in the meeting bar:

• "Reacquiring audio source" — capture is switching microphones; it clears on its own when audio resumes.
• "Couldn't save audio to disk — storage may be full" — Murmur couldn't write a chunk. Free up disk space; recording continues to try.

Crash-safe chunking

Murmur records in 5-minute chunks and writes each one to disk the instant it fills, as a compact M4A file. If the app crashes or your Mac restarts, at most the final in-progress chunk is at risk — every completed chunk is already saved and will be included when you recover or transcribe.

Stopping and transcribing

1. Click Stop. A confirmation dialog — Stop & Transcribe or Continue Recording — guards against stopping by accident.
2. The bar switches to a Transcribing… state (spinner + model name, no Stop button), so it's clear the recording has ended and processing has begun. (New in 1.6)
3. Your chunks are combined into a single playback file and transcribed in order; a progress indicator shows chunk X of Y.
4. When complete, the transcript appears and the meeting is saved to your Meetings list. If you enabled it, an AI summary is generated in the background.

Transcription time depends on recording length and the model you chose. A long meeting on a large model may take a few minutes; smaller models are significantly faster.

Recovering an interrupted meeting

(New in 1.6) The whole meeting pipeline is resumable. It runs in defined stages — combine → transcribe → summarize — and checkpoints after each, so an interruption never means starting over.

• Recovery sheet — when you open the Meetings tab after a crash or force-quit, Murmur detects any meeting that was mid-flight and offers to Recover or Discard it. Recovery picks up at the earliest unfinished stage: if audio was already combined but transcription hadn't finished, it resumes straight into transcription rather than redoing work.
• Scan disk for orphaned meetings — a button in Settings scans your meeting storage for recordings that have audio on disk but no library entry (for example, from a much older version) and recovers them. It's safe to run anytime: meetings already in your library are skipped, and it reports how many it found, recovered, and skipped.

Because your audio is chunked and saved as you go, and the pipeline resumes from its last checkpoint, a meeting you recorded is very hard to lose.

Choosing a transcription engine for meetings

All Murmur transcription engines work with meetings. A few things to consider:

• Parakeet — fastest for English, low latency, good for long recordings.
• WhisperKit — more accurate for non-English audio and accented speech; takes longer on large models.
• Apple Speech — good for English, very fast, but no speaker diarization support.