DictaWhisper docs

Quick start

pnpm dev

Open http://localhost:7777.

  1. Talk. Hit Record, or drag an audio file onto the page.
  2. Browser files begin processing immediately. Audio assessment/preparation runs before the selected transcription engine writes a sidecar .json. Bark Pro is the English default; Whisper is an alternative.
  3. If cleanup is configured, ollanet cleans the text and adds tags to the same JSON. The raw transcript is kept.
  4. Notes are grouped by month. Cleaned text is the default; raw speech is one click away. Playback follows the cleaned paragraphs.

If LocalSlip is installed, 7777 is the dictawhisper slip and 8008 is dictawhisper-api (loopback, not --lan).

Failure and recovery

What failed What you still have What to run
ollanet unreachable or not configured Recording plus raw text / segments Inbox Retry, or pnpm retranscribe --reclean when the host is up
Cleanup error on a note Same files. Status cleanup failed Retry on that note
Selected engine runtime missing Existing recordings and notes on disk. New transcription cannot finish Complete Bark setup, or fix whisper.python for a selected Whisper model, then run pnpm run doctor
ffmpeg or ffprobe missing Recordings and existing notes remain; audio preparation is unavailable Install both tools. Unfiltered mode still requires them

A healthy cleanup connection does not establish successful transcription. Check the selected engine’s result in Activity and use its recovery action when needed. Phone-folder sync stays optional and external.

Optional phone folder

If you also watch a phone folder (Syncthing or any folder sync), those files wait 30 minutes after the last write so the transfer can finish, then dated names are filed into YYYY/MM/ on that watch root.

Browser record/drop →  drop folder →  immediately ─────────────┐
Optional phone folder →  watch roots  →  settle 30m  →  YYYY/MM/ ┼─→ local transcription → sidecar.json → optional cleanup → UI
Force / retranscribe ─────────────────────────────────────────┘

Add the directory to "roots" in config.json:

"watch": {
  "roots": ["C:\\Users\\YOU\\VoiceNotes"]
}

Optional Tailscale

To reach the UI from a phone on your Tailscale:

"http": { "tailscale": true }

or DICTA_TAILSCALE=1. Doctor and startup print the MagicDNS / 100.x URL. There is still no login: only devices on that tailnet should be able to open it. The inbox is not bound to 0.0.0.0 or your LAN.

See Run for the other commands.