← Back to home
Updates
Changelog
-
v1.1.0 April 30, 2026 Latest The floating pill release.
A polish pass on the things you see most often: the moment you press the key, and the moment your text lands.
- Accuracy review (new, Pro flagship). A second-pass review reads your transcript like a careful editor and patches homophones, misheard proper nouns, dropped negations, and garbled technical words. Three modes: never, sometimes (only on recordings ≥ 60s, configurable), always. Preserves voice and structure — not a style rewrite.
- AI rewrite styles. Reword a transcript into formal, casual, concise, or email tone. Uses the chat model from your existing provider (Groq llama-3.3-70b or Gemini 2.0 Flash). No new vendor, no new key.
- Floating dictation pill. A small, glassy overlay appears at the bottom of your screen while you dictate, so you don't have to look away from what you're typing into. Hides itself when you release.
- Smart formatting pipeline. Word-boundary-aware voice commands, automatic capitalization, and a custom dictionary — shipped free to all users.
- Voice commands. Say "new line", "new paragraph", "period", "open quote", "bullet point", and more — punctuate without leaving the keyboard. See the full table.
- Custom dictionary. Add the names, jargon, and acronyms the model keeps mishearing. Per-user, stored locally.
- Multi-step onboarding wizard. First launch now walks you through welcome, provider and key, hotkey capture, and done.
- Reduced cold-start time by ~40% on Windows on ARM.
- Fixes: rare crash when releasing the trigger key during reconnect; settings window appearing off-screen on multi-monitor setups; tray icon ghosting after sleep.
-
v1.0.0 April 28, 2026 Initial release.
The first public version of Murmur. Everything you need to dictate; nothing you don't.
- Hold-to-talk dictation with any keyboard key, including unnamed media keys captured by scan code.
- Three modes: Smart (hold-or-tap-to-toggle), Hold (push-to-talk), and Toggle (tap to start, tap to stop).
- Two providers: Groq with whisper-large-v3 and Google Gemini, both bring-your-own-key.
- Real audio meter driven by your microphone level, not a faked animation.
- Local recents history with one-click clear.
- System tray presence with optional silent start, no console window.
- Windows 10 and 11 support, x64 and ARM64 builds.