Recording Quality
(New in 1.6) Murmur keeps a copy of your audio so you can play it back and re-transcribe it later. Settings > Audio lets you choose the quality of that saved audio, separately for each feature.
Two independent audio streams
It helps to know that Murmur runs two streams at once while recording:
- The transcription feed is always 16 kHz mono. That is exactly what the on-device speech models (Whisper, Parakeet) expect — they resample everything down to it internally, so a higher rate does not improve accuracy.
- The saved recording can be higher quality. (New in 1.6) Alongside transcription, Murmur captures a separate, native-rate stream and encodes it to a compact AAC file purely so your recordings sound clear when you listen back — clearer than the old 16 kHz "telephone-quality" playback.
This is why the quality setting affects playback only, and why it can never change your transcription results.
The settings
Each feature has its own Recording Quality section with two pickers:
- Sample Rate — 16 kHz, 24 kHz, 44.1 kHz, or 48 kHz (default).
- Bit Rate — 96, 128, 192, or 256 kbps. Default is 128 kbps for dictation and notes.
Higher values mean better-sounding playback and larger files. The default of 48 kHz · 128 kbps is a good balance for voice.
| Feature | Quality setting |
|---|---|
| Dictation | Sample rate + bit rate, default 48 kHz · 128 kbps |
| Notes | Sample rate + bit rate, default 48 kHz · 128 kbps |
| Meetings | Not shown — meeting audio is currently saved at 16 kHz (see below) |
Recording quality affects the audio saved for playback and re-transcription only. Transcription always uses 16 kHz mono, so higher rates do not improve accuracy.
Seeing the quality of a recording
In History, a recording's detail view shows a Recording Quality row, read from the actual saved file (for example "48 kHz · 128 kbps"). So you can always tell how a specific recording was captured, even if you have since changed the setting.
Meetings are a special case
Meeting audio is currently saved at 16 kHz mono. That is deliberate: the meetings chunking and crash-recovery pipeline is built around that format, and high-fidelity meeting playback is a separate future addition. The Meetings quality setting is therefore not shown, because it would have no effect today.
Related
- Microphone Priority — which mic capture uses
- Audio Playback — listening back to recordings