freestyle
Open source voice dictation platform for devs. Build voice workflows.
Freestyle is a desktop dictation app that types out what you say using a hotkey, running locally by default with optional cloud AI transcription providers.
Freestyle is a desktop dictation app for macOS, Windows, and Linux. The basic interaction is simple: hold a hotkey, say what you want to type, and release. The transcribed text appears wherever your cursor is, as if you had typed it. The README describes this as speaking four times faster than typing.
The app is designed to run locally, meaning your audio is processed on your own device and never sent to a third-party server without your explicit choice. That said, you can optionally connect it to external AI services by providing your own API key. Supported providers include OpenAI, Groq, Anthropic, Google, Deepgram, and ElevenLabs, giving you control over which transcription model handles your audio.
Beyond basic transcription, the app includes a few cleanup features. One removes filler words and fixes punctuation automatically after transcription. Another lets you define custom phrase substitutions, so if you frequently say something that gets transcribed incorrectly or in the wrong format, you can set a replacement rule. A third feature reformats your text based on context, so dictating into an email field produces email-formatted output rather than raw prose.
The project is available as a downloadable installer for each platform and is open to contributions through its GitHub repository. The license used is FSL-1.1-ALv2, which is a non-standard license; the README links to the full text but does not summarize its terms.
Where it fits
- Dictate text into any application by holding a hotkey and speaking
- Automatically remove filler words and fix punctuation after transcription
- Set custom phrase substitutions for words that transcribe incorrectly
- Get context aware formatting, like email style output when dictating into an email field