JustScribe changelog
JustScribe updates itself. Every version is also on GitHub Releases.
JustScribe v1.3.5
Dictation that knows your words, your apps, and when to stop
Four additions, all on your Mac as always.
- Vocabulary (Settings → Vocabulary). Add the names and words JustScribe must spell exactly — "Quassum", "SwiftUI", your colleagues — and, if you like, how the model tends to hear them. Misheard versions are fixed before the text is inserted, and ordinary words are left alone. Whisper models are also nudged towards your list. Works for file transcripts too.
- Press to start, press again to stop (Settings → Shortcut → Recording). Hold-to-record stays the default; in press mode the overlay says how to stop, a click on it stops too, and a forgotten recording ends by itself after 10 minutes.
- Voice commands (Settings → Behavior). Say "new line", "new paragraph", or "scratch that" to drop what you just said. In press mode, "stop recording" finishes, and "send" finishes and presses Return; in hold mode those are ordinary words. Spoken punctuation ("period", "comma", "open quote"…) is a separate switch, off by default.
- Clean-up modes (Settings → Clean-up, formerly Grammar Correction). Give the on-device clean-up model your own instructions — "formal, British spelling", "no filler words" — and bind modes to apps: the app you dictate into picks the mode, Default covers the rest, and a mode with clean-up off inserts your words as spoken (handy for a terminal).
Vocabulary and modes are stored in the app's own container; nothing leaves your Mac. The privacy policy at https://justscribe.quassum.com/privacy has the details.
JustScribe v1.3.4
Dictation history, if you want it
JustScribe can now keep a record of your dictations so you can find and reuse what you said. It is off by default: nothing is saved until you turn it on, and everything it keeps stays on your Mac, inside the app's own container.
- Settings → History has two switches: Keep Transcriptions saves the text of each dictation, with its date, length, model and language; Keep Audio Recordings also saves the recording as a small audio file (about 0.4 MB per minute). Audio is only kept when text is.
- History… in the menu-bar menu opens a window that lists your dictations, newest first, with search that ignores case and accents. Select one to read it in full, Copy it, Paste it straight back into the app you were in, Play its recording, or Delete it.
- Recordings are capped at 1 GB: once they pass that, the oldest recordings are removed and their text stays. Delete All History… in Settings removes everything.
- Turning a switch off keeps what already exists until you delete it.
The privacy policy at https://justscribe.quassum.com/privacy describes exactly what History stores.
JustScribe v1.3.3
Dictated text is pasted in one go
Until now JustScribe typed each word into the app you were in as it was recognised, then corrected it in place. Some text fields handle that stream of keystrokes badly. From this version the finished text is pasted once, when you release the key, after the final pass and any grammar correction.
- Insert Text By, a new choice in Settings → Behavior: Paste when done (the new default) or Type as you speak (the previous behaviour, for apps that intercept ⌘V).
- With Copy to Clipboard off, whatever was on your clipboard before — text, an image, a file — is put back after the paste.
- A dictation whose final pass comes back empty now keeps the words recognised while you spoke instead of reporting "No speech detected".
JustScribe v1.3.2
Transcribe audio and video files
A new Transcribe File… item in the menu-bar menu opens a window where you drop an audio or video file (m4a, mp3, wav, aiff, caf, flac, or the audio track of mp4, mov and m4v) and get a transcript with a timestamp on every paragraph. Copy it, or save it as a text file.
- Speaker labels. Turn on *Identify speakers* and each paragraph is marked Speaker 1, Speaker 2, … Leave the count empty to detect it, or enter the number of speakers. The first use downloads a small (about 22 MB) speaker model.
- Entirely on your Mac. Files are read in place, nothing is uploaded, and the transcript is discarded when you close the window unless you copy or save it. It uses the transcription model and language you already chose for dictation.
- Dictation comes first. If you dictate while a file is being transcribed, the file pauses and resumes when you let go.
- Works with Parakeet and Whisper models alike. Long files are processed a piece at a time, so memory stays flat.
Note: people moving from the Mac App Store version should install this release over the old copy; models and settings are kept, and macOS asks for Accessibility permission again.
JustScribe v1.3.1
- Link to privacy, terms, credits and support on justscribe.quassum.com
JustScribe v1.3.0
JustScribe moves off the Mac App Store
From this version JustScribe is distributed directly, as a signed and notarized download from https://justscribe.quassum.com, and updates itself.
- Automatic updates. The app checks for new versions daily and installs them in the background when it quits. Turn this off, or check now, in Settings → Updates.
- Apple Intelligence grammar correction. Optional grammar, spelling and punctuation correction can now use the language model built into macOS (no download) when Apple Intelligence is enabled. Llama 3.1 8B remains available as a downloadable alternative.
- Tip jar replaced by a link to sponsor the project on GitHub, since in-app tips only work on the App Store.
Moving from the App Store version: install this over the old copy; your models and settings are kept. macOS will ask for Accessibility permission again because the app is signed differently — if the switch has no effect, remove the old JustScribe entry under System Settings → Privacy & Security → Accessibility and add it again.