Native macOS dictation is unreliable—it frequently times out, fails to start, and struggles with technical jargon. Existing AI alternatives often demand $10+/month subscriptions just to use your own Mac's hardware.
Open Dictation is a headless utility designed to fix this.
- Zero Subscriptions: Run Whisper models locally on your Neural Engine or use your own API keys for cloud processing.
- Notch Integration: Visual feedback lives in the "Dynamic Island" area. No floating windows blocking your code.
- Developer Ready: Doesn't time out while you're thinking. Handles camelCase, file paths, and technical terms natively.
- No Bloat: 100% focused on getting text to your cursor. No sidebars, no proprietary "clouds," no nonsense.
- macOS 14+ (Sonoma or later)
- Apple Silicon Mac (M1, M2, M3, M4)
- MacBook with notch for the integrated UI (Dictation still works on non-notch Macs via audio cues).
-
Drag to Applications.
-
Open Terminal and run this to bypass the quarantine flag (required for non-signed builds):
xattr -rd com.apple.quarantine /Applications/OpenDictation.app
-
Launch and grant Accessibility and Microphone permissions.
- Option + Space: Start recording.
- Speak: The Notch will provide visual feedback while you talk.
- Option + Space: Stop recording. The transcribed text is inserted at your cursor instantly.
- Escape: Cancel the current recording.
| Feature | Description |
|---|---|
| Local Mode | Uses Whisper models on-device. No data leaves your Mac. |
| Cloud Mode | Connect to OpenAI, Groq, or any OpenAI-compatible API. |
| BYOK | "Bring Your Own Key"—pay the raw API cost ($0.003) instead of a platform markup. |
| Auto-Paste | Directly inserts text into the active field (IDE, Slack, Browser). |
The app automatically manages Whisper models stored in:
~/Library/Application Support/com.opendictation/Models/