Audio and transcripts
Recordings sit on a canvas as cards with a player and a transcript. A voice note captured on the phone becomes a task in one key, named from what was said, with the recording attached.
What you can do
- Paste or drop a recording on any canvas.
- Send a voice note from the iPhone app or the capture API: it lands on today's daily note with its transcript.
- Play it on the card; seeking works, Safari included.
- Open the folded transcript under the player.
- Make a task from a voice note: the name is the start of the transcript, the description is the whole transcript, and the recording moves to the task.
- Have the server transcribe recordings, where the server has transcription configured.
How to
- Drop an audio file on the canvas, or paste it with Mod+V.
- Press Transcript on the card to unfold it.
- Select the card and press Shift+T to make it a task, or right-click it.
Limits
- Types: M4A, AAC, MP3, WAV and WebM, read from the file itself. Up to 25 MB.
- Server transcription is off unless the server is configured for it (any OpenAI-compatible transcription service). Without it, the transcript is the one the device made, or none.
- A server transcript replaces the device one, unless the transcript was edited first. If the server fails, the device transcript stays. The card shows Transcribing while it runs and "Server transcript failed" if it gave up.
- A task made while a server transcript is pending keeps the transcript as it was.
- A voice note with no transcript becomes a task named "Voice note" with the date and time.
- Task names take up to 120 characters of the transcript, cut on a word.
- Audio cards do not become docs.
<!-- verify: how the transcript is edited on the card (canvas.md says editable like a note's) -->