OpenWhispr
Desktop dictation that types where your cursor is, using local or cloud speech models.
- Category
- Content & Media
- Website
- github.com
- Language
- English
- Listed
- 11d ago
- Stars
- 8,321
- Forks
- 1,025
OpenWhispr/openwhispr ↗, JavaScript, MIT, Last commit 1d ago
Press the hotkey, speak, and the transcribed text lands wherever your cursor is. Transcription runs either on-device through Whisper or NVIDIA Parakeet, or through a cloud model with your own API key. Beyond dictation the app records and transcribes meetings, keeps notes, and exposes a chat agent over the same audio. There is no data collection and no telemetry.
Builds for macOS, Windows, and Linux come from the releases page under an MIT license. Intel Macs lose live speaker identification and voice fingerprinting, because both depend on ONNX Runtime, which stopped shipping macOS x86_64 binaries in version 1.24.