Step 1
Activate
Hold a key or use hands-free mode.
Speak freely.
It never leaves your device.
AllUSay is a dictation app for the Mac. Speech recognition runs on your device and your words are typed into the app you're using.
Offline edition: no internet needed, the speech model is included.
Version 0.6.2 · 50 MB disk image (offline edition 328 MB) · Requires an Apple Silicon Mac and macOS 14 or later.
Works in the apps you already use.
Activate. Speak. Release.
Step 2
Speak
Talk normally.
Step 3
Release
AllUSay types clean text where your cursor is.
Your data never leaves your device.
Your data stays on your Mac.
- No sign-up needed.
- No analytics or any form of tracking.
- Works fully offline.
- Transcribed text stays on your device, as long as you want to keep it.
When your words can't leave the room.
Offline and isolated machines
Works fully offline: On isolated machines, at secure sites or on a plane.
Confidential work
Client notes, case files, patient letters, HR reviews.
When cloud services are off-limits
No tokens to track or internet permissions to request.
Private writing
Journals, personal notes, letters you're not ready to send.
Questions, answered.
Which languages are supported?
English, German, French, Spanish, Italian, Dutch, Portuguese and Polish. Language is auto-detected. Filler removal and spoken corrections work in English, German and French.
What's the difference between standard and offline?
Standard version is ~50 MB, downloads the model once and auto updates. Offline version is about 330 MB, comes with everything included and doesn't auto update.
How long can a single dictation be?
Up to 15 minutes.
Can I move my Dictionary and Snippets to another Mac?
Yes. Both can be exported and imported.
What runs on the device?
Speech recognition: TranscribePsy-250M-EU-sub by QVAC (305 MB), fine-tuned from NVIDIA Parakeet TDT 0.6B v3.
How accurate is it?
Word error rate on public datasets varies between 4% to 8%.
How fast is it?
1 sentence: ~150 ms. 60 words: ~500 ms. p95 across test recordings: ~250 ms. 15-minute dictation: ~12 s.
What are the memory and disk requirements?
About 600 MB of memory while running, mostly for the speech model. The download is about 50 MB (the app takes about 130 MB once installed) and the speech model 305 MB. The offline edition is ~330 MB, with the model included.
What permissions does the app require?
Microphone, to hear your words, and Accessibility, to type into other apps.