Help · We read everything
Support
Something not working, or a question that isn't answered below? Get in touch, and a real person reads every message.
Email: support@visionspeech.app. When reporting a problem, it helps to mention your device, its OS version (macOS, iOS, or visionOS), and what you were doing when it happened.
Frequently asked
How do I dictate?
It depends on the platform. On Mac, hold the dictation key (Right Option by default), speak, and release — the transcript pastes at your cursor in the frontmost app; double-tap Command for hands-free recording. On iPhone, tap the mic on the VisionSpeech keyboard from any text field; the finished transcript inserts itself. On Apple Vision Pro, tap the orb (or the record button in Studio) to start and again to stop; the transcript is copied to your clipboard, ready to paste.
Which languages are supported?
You choose between two models, both running on the device. Parakeet v3 detects 25 European languages on its own, so you just speak, even switching languages partway through a sentence. Whisper Large v3 covers around 100 languages; let it detect the language or pick a specific one. Choose your model when you set up, and change it any time in Settings.
Why does it download a few hundred MB the first time?
That is the speech model you chose: about 470 MB for Parakeet v3 or 626 MB for Whisper Large v3. It downloads once, then transcription runs entirely on your device, with no network needed afterward, and nothing you say is ever uploaded. Switching models later downloads that model once, too.
The model download failed or got stuck.
Make sure you have a stable internet connection with room to spare for the model (up to ~626 MB), then retry from Settings. If it still fails, quit and reopen the app, or delete and reinstall. There are no accounts, so you lose nothing.
The app cannot hear me.
VisionSpeech needs microphone access. On Mac, check System Settings ▸ Privacy & Security ▸ Microphone; the Mac app also has its own microphone picker in Settings ▸ General if you use an external mic. On iPhone and Vision Pro, check Settings ▸ Privacy & Security ▸ Microphone and confirm VisionSpeech is enabled, then try recording again.
How do I set up the Mac hotkeys and pasting?
The Mac app needs the Accessibility permission once (System Settings ▸ Privacy & Security ▸ Accessibility) — it covers both the global hold-to-talk hotkey and pasting the transcript into the app you are writing in. Without it, VisionSpeech still records and copies the transcript to your clipboard. Note that hotkeys go quiet while a password field is focused; that is a macOS security behavior, not a bug.
How do I enable the iPhone keyboard?
Add it in Settings ▸ General ▸ Keyboard ▸ Keyboards ▸ Add New Keyboard ▸ VisionSpeech, then switch to it with the globe key. When you tap its mic, the VisionSpeech app opens briefly to start a dictation session — keyboards cannot use the microphone themselves, and transcription runs in the app, on the device. Swipe back to where you were typing; the finished transcript inserts itself.
Can it paste directly into other apps?
On Mac, yes — with the Accessibility permission granted, the transcript pastes at your cursor in the frontmost app, and your previous clipboard is restored afterward. On iPhone, the VisionSpeech keyboard types the transcript into whatever field you are in. visionOS does not allow apps to type into other apps, so on Vision Pro the transcript is copied to your clipboard and you paste it in one gesture.
It keeps misspelling a name or technical term.
Teach it the word. Open My Words and add the spelling once — names, brands, jargon — and transcripts will snap close matches to it from then on. You can also enable a bundled dictionary pack for your field. For phrases that should always come out a specific way (like a spoken email address turning into the real one), add a Text Replacement instead.
How does AI cleanup work, and why don’t I see it?
AI cleanup can remove filler words, fix punctuation, or reformat the text as bullet points or email tone, and one tap reverts to the original. It runs on Apple Intelligence, entirely on the device; on Mac, languages Apple Intelligence does not cover can use a separately downloaded local model instead (Settings ▸ Models). On devices where neither is available, the option hides or explains itself.
Why is Whisper slow the first time after opening the app?
Whisper Large v3 is a big model, and the first preparation after each app launch can take a few minutes while it loads into memory. Keep the app open; once prepared, it stays fast until the app closes. Parakeet is usually ready in under a minute after download.
Can I start dictation without opening the app?
Yes. On Mac, the global hotkeys work from anywhere. Everywhere, you can say "Hey Siri, start dictation", and on iPhone and Vision Pro you can pin the launcher widget and tap it — the app opens and starts recording immediately. There are also Stop Dictation and Get Latest Transcript actions for Shortcuts, so you can pipe your last transcript into other workflows.
Where are my past transcripts?
Your last 100 transcripts are kept locally on the device. Open the history to revisit, edit, copy, pin, share, or delete them. Pinned transcripts survive the 100-item limit. On iPhone and Vision Pro you can optionally turn on iCloud Sync to mirror history through your iCloud account; the Mac keeps everything local. Clearing history or deleting the app removes local copies for good.
What do you require?
A Mac on macOS 26, an iPhone on iOS 26, or an Apple Vision Pro on visionOS 26, or later. AI cleanup additionally requires Apple Intelligence (or, on Mac, the downloadable local model).
Privacy
VisionSpeech processes your speech on the device and collects no data. For the full details, read the privacy policy.