Skip to main content

​
Dictate with the hotkey

Freestyle picks a sensible default hotkey per platform: Dictation defaults to Hold (push-to-talk). There are two activation modes:

Hold (push-to-talk)

Hold to record, release to transcribe. Great for quick bursts. This is the default.

Toggle

Press once to start, press again to stop. Great for longer dictation.
To change it, go to Settings → Recording. Pick Hold or Toggle under Activation, and click the current shortcut to rebind it (it needs a modifier key or a side mouse button).
On Linux, true hold-to-talk needs access to the input device. If it isn’t available, Freestyle falls back to toggle mode and tells you how to enable it.
Under Output mode you can choose Paste into app (text is typed straight into the active window) or Copy to clipboard (you paste it yourself). Clipboard mode helps with apps where auto-paste misbehaves.
On macOS, paste needs Accessibility permission. On Linux it uses xdotool or wtype. If nothing shows up, check Settings → Permissions.

​
Work with Remix

Remix is Freestyle’s agent workspace. It keeps the conversation in the center of the desktop window so you can ask a question, continue a thread, or work from a selection without losing your place. Use the Remix hotkey to begin a voice request: Fn + Control on macOS and Ctrl + Alt + E on Windows and Linux. You can change it under Settings → Remix. A short press opens Remix; hold it while speaking to send a request. Use the expand control in the floating Remix surface to continue in the full workspace. In the workspace, the Context control opens a right rail with Tasks, Notes, and Brain. These are quick context surfaces, not separate pages: you can add or complete a task, read a note or file preview, and expand a card to see more while the chat stays visible. Hide the rail when you need more room for the conversation. For tools beyond the built-in connected apps, add a trusted local or remote server at Settings → Remix → MCP connections. OAuth, tokens, and local commands stay in the desktop app. Remix Cloud sees the enabled tool descriptions it needs to request a call, then receives only that call’s bounded, sanitized result so it can complete the reply.
Models and connected apps are all inside Settings. The app does not open a separate Settings window.

​
Transcription

Freestyle Transcribe is the built-in transcription service, built by the Freestyle team and tuned for developer dictation. Sign in and it works — there is nothing to configure and no API key to paste.
  • Transcription and cleanup in one pass, and fast.
  • Zero-day retention by default. Transcriptions are private and not stored.

​
Shape the final text

Freestyle has four features that turn raw speech into exactly the text you want. Each has its own page with examples and details:

Remix

Ask, write, and research from the agent workspace without losing your chat context.

AI cleanup

Run your transcript through a language model to fix punctuation, drop filler, and polish phrasing — at the intensity you choose.

Dictionary

Swap spoken phrases for exact text after transcription, like type script → TypeScript.

Vocabulary

Teach the speech model tricky names, brands, and jargon so it spells them right.

Formats

Reshape text based on the app you’re in — an email in Gmail, a casual note in Slack.

​
Next steps

Build a plugin

Extend the dictation pipeline with your own code.

Join the community

Ask questions and hang out with contributors on Discord.