BetterWispr
Free, open-source voice dictation for macOS. Hold ⌥ Space, speak, release, and the text lands in whatever app you were typing in.
BetterWispr is a private, offline speech-to-text app and an open-source alternative to Wispr Flow. It transcribes on your Mac with Whisper, NVIDIA Parakeet or Apple's on-device speech recognition, so your voice never leaves your computer unless you add an API connection yourself.
Features
- Dictate anywhere: hold-to-talk or press-to-toggle, with a shortcut you choose.
- On-device models: Apple speech, Parakeet v3 (25 European languages), Parakeet v2 (English) and Whisper large-v3 turbo.
- Vocabulary: spelling hints and phrase replacements for names and jargon.
- Learns your corrections: fix a misheard word once (in History, or right after it's pasted) and it's spelled right next time.
- Voice commands: say "comma", "new line", "scratch that" or "sorry, remove that", and "at the rate KV" for "@KV".
- Meetings: live transcript of you and the other side, plus on-device titles, summaries and action items with Apple Intelligence or a local Ollama model.
- History: raw and corrected transcripts, stored locally (can be turned off).
- Bring your own model: optional Sarvam AI, Smallest AI or any OpenAI-compatible endpoint.
Requirements
- macOS 14 or later
- Apple Silicon recommended
Getting started
- Download the DMG from Releases and drag BetterWispr to Applications.
- Open it and follow the welcome guide.
- Pick a model in Models. Apple on-device needs no download; Parakeet and Whisper download once, then work offline.
- Grant Microphone access, and Accessibility access so BetterWispr can paste for you.
- Click into a text field in any app, hold ⌥ Space, speak, and release.
If paste can't safely reach the previous app, the text is copied to your clipboard instead, so you never lose a dictation. Recordings stop automatically after two minutes.
Bring your own model
Models → Bring your own model → Add connection. Choose a provider, enter the model ID and API key, save, then click Use.
| Provider | Default model | Notes |
|---|---|---|
| Sarvam AI | saaras:v4 |
Long recordings are sent in 25-second chunks |
| Smallest AI | pulse |
Set an explicit spoken language in Settings |
| OpenAI-compatible | any | Full URL, e.g. http://localhost:8000/v1/audio/transcriptions. HTTPS required except on localhost |
When a connection is in use, your audio goes to that endpoint and its pricing and retention policies apply. Keys are stored in the macOS Keychain. Switch back to a built-in model to return to fully local transcription. BetterWispr never falls back to a cloud provider on its own.
Privacy
- Settings, vocabulary and history live in
~/Library/Application Support/BetterWispr/workspace.json(not encrypted). - Models live in the
Modelsfolder next to it. - Audio is written to temporary files and deleted after each recording.
- Built-in models only touch the network when you explicitly download one.
Troubleshooting
- Paste stopped working after an update or rebuild: remove and re-add BetterWispr in System Settings → Privacy & Security → Accessibility.
- Apple on-device model unavailable: your language or OS speech assets may not support it. Install Parakeet or Whisper instead.
- First transcription is slow: large models take time to load the first time.
Contributing
Build instructions, project layout and guidelines are in CONTRIBUTING.md.
License
Apache-2.0. Third-party code and model weights keep their own terms;