Fast, local-first voice app for Mac. Free and open-source.
There are many voice transcription/dictation apps, but this one is mine.
MacParakeet runs NVIDIA's Parakeet TDT on Apple's Neural Engine via FluidAudio CoreML. Press a hotkey, speak, text appears. Or drag a file and get a full transcript. All speech recognition happens on your Mac.
Dictation — Press a hotkey in any app, speak, text gets pasted. Hold for push-to-talk, double-tap for persistent recording. Works system-wide.
File transcription — Drag audio or video files, or paste a YouTube URL. Full transcript with word-level timestamps, speaker labels, and export to 7 formats (TXT, Markdown, SRT, VTT, DOCX, PDF, JSON).
Text cleanup — Filler word removal, custom word replacements, text snippets with triggers. Deterministic pipeline, no LLM needed.
AI features — Optional transcript summarization and chat via your own API keys (OpenAI, Anthropic, Ollama, OpenRouter). Entirely opt-in.
- ~155x realtime — 60 min of audio in ~23 seconds
- ~2.5% word error rate (Parakeet TDT 0.6B-v3)
- ~66 MB working memory during inference
- 25 European languages with auto-detection
- Apple Silicon only (M1/M2/M3/M4)
- Best with English — supports 25 European languages but accuracy varies
- No CJK language support (Korean, Japanese, Chinese, etc.)
Download: Grab the notarized DMG or visit macparakeet.com. Drag to Applications, done.
First launch downloads the speech model (~6 GB). After that, dictation and transcription work fully offline.
Build from source:
git clone https://github.com/moona3k/macparakeet.git
cd macparakeet
swift test # 976 tests
scripts/dev/run_app.sh # build, sign, launchThe dev script creates a signed .app bundle so macOS grants mic and accessibility permissions. Set DEVELOPMENT_TEAM=YOUR_TEAM_ID if needed.
CLI:
swift run macparakeet-cli transcribe /path/to/audio.mp3
swift run macparakeet-cli models status
swift run macparakeet-cli history| Layer | Choice |
|---|---|
| STT | Parakeet TDT 0.6B-v3 via FluidAudio CoreML (Neural Engine) |
| Language | Swift 6.0 + SwiftUI |
| Database | SQLite via GRDB |
| Auto-updates | Sparkle 2 |
| YouTube | yt-dlp |
| Platform | macOS 14.2+, Apple Silicon |
All speech recognition runs on the Neural Engine. Your audio never leaves your Mac.
- No cloud STT. The model runs on-device. No audio is transmitted.
- No accounts. No login, no email, no registration.
- Anonymous telemetry. Non-identifying usage analytics, opt-out in Settings. No persistent IDs, no IP storage, no content transmitted. Source code is right here — verify it yourself.
- Temp files cleaned up. Audio deleted after transcription unless you save it.
What does use the network: AI Summary & Chat connects to LLM providers when you configure it with your own API keys. YouTube transcription downloads video via yt-dlp. Telemetry pings our server unless you opt out. Core dictation and transcription are fully offline.
Note: Builds from source also send telemetry by default. Opt out in Settings or set MACPARAKEET_TELEMETRY_URL to override.
- Report bugs — Open an issue
- Submit a PR — Fork, make changes,
swift test, open a PR - Read the specs — Architecture decisions and feature specs live in
spec/
For larger changes, open an issue first.
GPL-3.0. Free software. Full license.
Made for people who think faster than they type.
