Chirp

Free, Fast, Offline voice-to-text for macOS. Hold fn, speak, release — text appears at your cursor. All transcription runs on-device — no accounts, no servers, no data leaves your machine.

Install

Download

Grab the latest DMG from GitHub Releases, open it, and drag Chirp to Applications.

Homebrew

brew install --cask stefanpenner/chirp/chirp

Build from source

brew install bazelisk    # install Bazel via Bazelisk
bazel run //:Chirp       # build and launch

Features

• Hold-to-talk — Press fn (configurable) to record, release to transcribe, hotkey can be changed
• Works everywhere — Text is typed at your cursor in any app
• Speculative preview — See partial transcription as you speak
• Fast — Parakeet TDT 0.6b v2 with Silero VAD, all on-device
• Menu bar app — No dock icon, no clutter
• Auto-updates — Check for updates from the menu bar

Setup

Permissions

Chirp needs two permissions on first launch:

1. Microphone — macOS will prompt automatically. Click Allow.
2. Accessibility — Required for typing text at your cursor. Go to:
   • System Settings → Privacy & Security → Accessibility
   • Click the + button, find Chirp in Applications, and add it
   • Make sure the toggle is enabled

If Chirp can transcribe but text doesn't appear, Accessibility permission is missing.

Model download

On first launch, Chirp downloads the Parakeet TDT 0.6b v2 speech recognition model (~460 MB download, ~631 MB on disk). This is a one-time download stored in ~/Library/Application Support/Chirp/. After that, Chirp works fully offline.

Updating

Chirp checks for updates automatically. You can also check manually from the menu bar icon → Check for Updates…

Or with Homebrew:

brew upgrade chirp

Troubleshooting

Text doesn't appear after transcription

Accessibility permission isn't enabled. Go to System Settings → Privacy & Security → Accessibility and make sure Chirp is listed and toggled on. You may need to remove and re-add it if you moved the app.

No audio detected / microphone not working

• Check System Settings → Privacy & Security → Microphone and ensure Chirp is allowed
• Make sure your input device is set correctly in System Settings → Sound → Input

Model download fails

Chirp needs internet access for the initial model download. If it fails, check your network connection and restart the app. The download resumes where it left off.

Requirements

• macOS 26+
• ~700 MB disk space (app + model)
• Microphone permission
• Accessibility permission

Privacy

All transcription runs locally. Your audio is processed in memory and never recorded or stored. No telemetry, no tracking, no data leaves your machine. Other than downloading the open-source models, Chirp requires no network access.

Acknowledgments

Chirp is built on these open-source projects:

• sherpa-onnx — speech recognition engine (Apache 2.0)
• ONNX Runtime — model inference runtime (MIT)
• NVIDIA Parakeet TDT 0.6b v2 — speech-to-text model (CC-BY-4.0)
• Silero VAD — voice activity detection (MIT)
• Sparkle — auto-update framework (MIT)

Full license texts are in THIRD_PARTY_NOTICES.

Development

Install Bazelisk, which reads .bazelversion and fetches the right Bazel automatically:

brew install bazelisk
bazel build //:Chirp                  # build
bazel run //:Chirp                    # build and launch
bazel test //...                      # run tests
bazel run //:package -- 0.3.0         # create signed DMG