Re-transcription

Murmur keeps the original audio for every dictation, note, and meeting. That means a transcription is never final — you can re-run any recording through a different (or better) model later.

Your audio is saved before transcription runs. (New in 1.6) Murmur writes the recording to disk and creates its History entry the moment you finish speaking — before any transcription happens. So if transcription fails, or the app quits mid-process, the audio is still there and the recording still appears in History, ready to re-transcribe. You won't lose a recording to a transcription error.

Re-transcribing a single recording

1. Open History and select the recording.
2. Click Re-transcribe in the detail view.
3. The audio is run through your currently active engine, and the text, model name, and timestamps are updated.

This is useful when:

• A transcription came out poorly and you have since downloaded a more accurate model
• You originally transcribed in the wrong language setting
• A transcription failed or was left blank

Re-transcription needs the original audio file. Imported files that were transcribed without keeping a copy cannot be re-run.

Batch re-transcription

When many recordings are pending — typically after library recovery, or after a string of failures — process them all at once:

1. Open Settings > Developer.
2. Click Re-transcribe Pending Recordings — "Run every pending recording through the active engine, one at a time."
3. Review the pending count and total audio duration, then click Start.

While running, the view shows a progress bar, the file currently being processed, an estimated time remaining (after the first few items), and live counts of succeeded, failed, and skipped items. It is safe to minimize the window — just keep Murmur running until it finishes.

Stopping and resuming

Click Stop at any time. Everything processed before the stop is saved; rerunning the batch later picks up where you left off. If nothing is pending, the view simply reports "Nothing pending — you're all caught up."

After the batch

A summary lists attempted, succeeded, failed, and skipped counts. Failed items remain in History where you can retry them individually or include them in another batch run. A fatal error never touches transcriptions that already completed.

Which engine is used

Both single and batch re-transcription use the active model selected in Settings > Dictation Models. To re-transcribe with a more accurate model, switch the active model first — see Downloading & Managing Models.