Release notes
Changelog
Polish you can hear.
Small details, big feel: gentle audio cues, an escape hatch for a hot mic, and the privacy features we promised, now in the app.
New
- Soft audio cues. A gentle chime up when recording starts, down when it stops, and a low blip if something goes wrong. Synthesized in-app, quiet by design, and easy to switch off in Settings.
- Press Esc to cancel. While the mic is hot, Esc discards the recording: nothing is transcribed, nothing is pasted.
- Privacy Mode. Flip it on and dictation history never touches disk. Aggregate stats (counts only, never content) still work.
- Snippets. Say a trigger phrase ("insert my sign-off") and a saved block pastes in its place. Define them in Settings · Vocabulary.
- Smarter command mode. "Make this a bulleted list" and friends now reshape the dictation instead of being typed out.
Fixed
- Canceling a dictation now reliably discards the audio instead of pasting it anyway.
- A microphone error mid-recording now resets the session cleanly.
- Overlay status messages no longer vanish early (or linger) when dictations overlap.
- "Your rhythm" insights on the Voice Profile now compute correctly.
- Clearer, calmer error messages throughout.
Hello, Whurd.
The first public build. Press a hotkey, talk into any Windows app, and get clean text back: filler removed, punctuation fixed, styled for where you're writing.
New
- System-wide dictation. Trigger from anywhere with Ctrl+Shift+Space. The dictation HUD never steals focus from your active app.
- Instant transcription. Speech-to-text that returns in roughly 300ms.
- AI formatting. Automatic filler removal, punctuation, capitalization, and sentence repair so the text reads like you wrote it on purpose.
- Per-app auto-styles. Output adapts to the active app: crisp for Slack, formal for email, terse for a commit message.
- Custom dictionary. Teach Whurd your names, jargon, and code symbols so they spell correctly every time.
- Command mode. Say "make this a bulleted list" and have the text restructure on the fly.
- Voice Profile analytics. Track words dictated, time saved, and WPM over time, with achievements. Stored locally.
- 99 languages. Dictate in the language you think in.
- Privacy Mode. Keep dictation history off-disk entirely, available on the free tier.
Privacy
- Voice audio is used only for transcription, then discarded. Never stored, never used to train.
- API keys and dictation history are stored locally on your machine.
- AI-training opt-in ships off by default.
Known limitations
- Windows 10 and 11 only for this release.
- Per-app styles cover the most common apps; more profiles land in upcoming builds.