[ RELEASE NOTES · CHANGELOG ]
what's new.
macOS 1.0.12 — 2026-09-10
- Fixed: some dictations pasted nothing, or pasted your previous clipboard. If another app was holding the clipboard at the moment blah. tried to use it — a clipboard manager, or a remote desktop session syncing in the background — the text never made it there, but blah. pressed paste anyway. You got nothing, or, with "restore previous clipboard" turned on, whatever you had copied before. blah. now checks that your dictation actually reached the clipboard before pasting, waits longer before putting your old clipboard back, and tells you in the menu if the clipboard was unavailable instead of failing quietly.
- Fixed: dictation didn't work for other accounts on a shared Mac. If one person installed blah. and someone else used it from their own account on the same Mac, the speech model couldn't be opened and nothing was transcribed. It now works from every account.
macOS 1.0.9 — 2026-08-12
- Live captions now work with the privacy filter on. Captions used to switch themselves off whenever the filter was enabled — silently, so the feature just looked broken. Now they stay on. What gets pasted (and what goes into stats and transcript history) is still filtered; the on-screen captions of words you just said out loud are not.
- blah. now tells you when voice modes can't run. If the language model can't load, your dictation still pastes — but you get a clear warning in the menu instead of quietly unstyled text.
- "Run setup again" no longer switches your voice back on. If you'd deliberately set Voice to Off, re-running setup used to flip it back to Simple Rephrase when you clicked through.
- Now requires macOS 15 or later.
macOS 1.0.8 — 2026-07-28
- Fixed: voice modes weren't transforming your text. A bad bundled model meant dictations with a voice selected (Formal, HR, Bullets, etc.) pasted the raw transcription instead of the styled result. Voices work again.
- Fixed: dictating into a remote desktop pasted the previous dictation. When pasting into a remote-desktop or virtual-machine window (Microsoft Windows App, Parallels, VMware Fusion, Citrix, VNC, and friends), blah. now gives the remote connection time to sync your text before pasting — so each dictation pastes what you just said. Remote pastes take about half a second longer; everything local is unchanged.
- Clipboard safety everywhere: blah. no longer overwrites something you copied while a dictation was still pasting.
macOS 1.0.7 — 2026-07-12
The auto-update release. This is the last version you'll ever have to install by hand.
- Automatic updates. About blah. → Check for Updates… finds new versions, downloads them, and installs in place — updates are tiny and take seconds, not a full re-download.
- Smarter cleanup. Voice cleanup and translation now stick to transforming your words — dictating things like "ignore that" or questions no longer derails the output.
- Updates check only when you click the button — blah. never phones home on its own.
macOS 1.0.3 — 2026-07-08
Activation, smoothed out.
- Clearer messages when a key can't activate — "this license is active on another computer" and "this license moved recently — try again in N days" are now distinct, so you know exactly what to do.
- Activation handles rare Mac hardware hiccups safely instead of leaving your license in a confusing state.
macOS 1.0.1 — 2026-07-07
First public beta.
- Download, drag to Applications, launch — no security warnings, no hoops.
- Everything in one app: local dictation, 30 voice modes in 5 groups, translate (21 languages), privacy filter, live captions, voice memos, custom vocabulary + macros, shift-hold preview editing, file transcription, transcript history (opt-in), stats + desktop widget.
- Fully local: your voice and your text never leave your Mac.
Windows 1.0.2 — 2026-08-17
The first public Windows release.
- Dictate anywhere. Hold your hotkey, speak, release — your words paste into whatever app you're working in. Whisper runs entirely on your own GPU; nothing you say ever leaves your computer.
- Voices. Thirty ways to shape the output — clean up filler, bullet lists, formal tone, code in eight languages, or just pirate — all running locally.
- Translate as you dictate. Speak English, paste any of 21 languages.
- Privacy filter. Flag or mask names, emails, phone numbers, cards and addresses before they're pasted — detection runs locally, like everything else.
- Live captions, voice memos, file transcription. Watch words appear as you speak, capture quick audio notes, or drop an audio/video file for a transcript.
- Check for updates from the app. About → "Check for Updates…" downloads and installs new versions in place — updates are small and never re-download the speech models.