Whisnap

---

name: whisnap description: macOS CLI for transcribing audio and video files using local Whisper models or Whisnap Cloud. homepage: https://whisnap.com metadata: {"clawdbot":{"emoji":"🎙️","requires":{"bins":["whisnap"]},"install":[{"id":"app","kind":"manual","label":"Install via Whisnap app Settings → Advanced → Enable CLI"}]}}

---

whisnap

Use whisnap for transcribing audio/video files from the terminal. Requires the Whisnap macOS app with at least one model downloaded.

Setup (once)

Open Whisnap app → Settings → Advanced → Enable CLI (creates /usr/local/bin/whisnap symlink)

Download at least one Whisper model in the app

Common commands

Transcribe audio: whisnap recording.wav

Transcribe video: whisnap meeting.mp4

Cloud transcription: whisnap recording.wav --cloud

JSON output with timestamps: whisnap lecture.m4a --json

Specific model: whisnap interview.wav -m small-q5_1

Cloud + JSON: whisnap recording.wav --cloud --json

List downloaded models: whisnap --list-models

Verbose diagnostics: whisnap recording.wav -v

Supported formats

Audio: WAV, MP3, FLAC, M4A, OGG

Video: MP4, MOV, MKV, WebM

Flags

-c, --cloud — Use Whisnap Cloud instead of local model (requires sign-in)

-m, --model — Override model (e.g., small-q5_1). Defaults to app's selected model.

-j, --json — Structured JSON output with text, segments, timestamps, model info

-v, --verbose — Print progress and diagnostics to stderr

--list-models — List available models and exit

JSON output format

{
  "text": "transcribed text",
  "segments": [{ "start_ms": 0, "end_ms": 1000, "text": "segment" }],
  "model": "small-q5_1",
  "backend": "whisper",
  "processing_time_ms": 5000
}

Notes

The CLI reuses models and settings from the Whisnap app (~/Library/Application Support/com.whisnap.desktop/).

Cloud mode requires authentication — sign in via the app first.

For scripting, use --json and pipe stdout. Diagnostics go to stderr.

Exit code 0 = success, 1 = error.

Only Whisper models are supported in CLI mode (not Parakeet).

Confirm the file path exists before transcribing — the CLI validates but does not search.