Privacy Policy
This policy covers the app. Website visits, purchases and support reports have separate data handling. If you send a report, its message and attachments leave your Mac so we can respond. Direct license activation uses Lemon Squeezy.
Last updated: September 2026
The short version: WhisperType collects nothing. There's no account, no analytics, no advertising, and no servers of ours in the loop. Your voice is captured in memory, transcribed on your Mac's own Neural Engine, and never written to disk or sent anywhere. This page explains exactly what that means.
What happens to your audio
When you dictate, WhisperType captures microphone audio into memory, converts it to text on-device using a Whisper model running on the Apple Neural Engine, and then discards the audio. Nothing is ever written to disk, uploaded, or synced. The transcription is placed into the app you're using (and on your clipboard) and lives only where you put it.
The transcription, and your clipboard
- ✓The text is inserted into whatever field currently has focus, exactly as if you'd typed it — it stays in that app, on your Mac.
- ✓To paste, WhisperType briefly uses the clipboard, then restores your previous clipboard contents — your copied data isn't kept or sent.
- ✓Prefer to paste yourself? Turn off auto-paste and the text simply waits on your clipboard for a manual ⌘V.
The live preview
While you are still speaking, WhisperType shows a rough preview of your words in the floating panel and the Studio. That preview comes from macOS's own speech recogniser, not from us and not from a server: audio is handed to a system service on your Mac and the words come back locally.
Apple's recogniser is capable of sending audio to Apple's servers, so WhisperType does not leave that to chance. Every request it makes is pinned to on-device recognition, and if a Mac cannot do it offline — some Intel Macs need the dictation model downloaded first — the preview simply does not appear. There is no network fallback: a missing preview is acceptable, sending your voice anywhere is not. An automated test fails the build if that pin is ever removed.
The preview is on by default. The first time it runs, macOS asks for its Speech Recognition permission, separate from the microphone — decline it, or switch the preview off in the Control Panel, and dictation works exactly as before. What the preview shows is never what gets inserted — the text that lands in your document always comes from Whisper.
On-device cleanup and per-app modes
If you turn on AI cleanup (macOS 26 and later), WhisperType hands the transcribed text to the language model that is already part of macOS, through Apple’s FoundationModels framework. That model runs on your Mac. Your words are not sent to Apple, to us, or to anyone else, and the feature is off until you switch it on. Custom vocabulary, text replacements and per-app modes are plain settings stored locally and applied on your Mac.
What's stored, and where
- ✓Preferences (your hotkey, model choice, and options) are kept in standard macOS app preferences, local to your Mac.
- ✓The AI model files are downloaded once into your Documents folder (
~/Documents/huggingface/…) and used offline from then on. - ✓Your licence and trial dates live in your Mac’s Keychain, so deleting preferences doesn’t lose a licence you paid for. If you bought from the Mac App Store instead, the purchase is validated through StoreKit under your own Apple ID. Either way, WhisperType stores no payment details.
- ✓Your free-plan counter — how many seconds of speech you have transcribed today — is a single number in the Keychain, replaced each day and never sent anywhere. It records a duration, not what you said.
- ✓Your dictation history (the last 100 transcriptions) and your usage statistics stay in local app preferences on your Mac. Clearing the history removes them; nothing was copied elsewhere.
The few things that touch the network
Downloading a model. The first time you use a model you don’t have yet, WhisperType downloads it from the public Hugging Face repository argmaxinc/whisperkit-coreml. That request contains no personal data — it only fetches the model files. Once it completes, transcription works entirely offline, with Wi-Fi off if you like.
Activating a licence. If you bought directly rather than from the App Store, activating your key sends the key and the name of your Mac (the name in System Settings ▸ General ▸ About, used so you can tell your activations apart) to our payment provider, Lemon Squeezy, at api.lemonsqueezy.com. That is the only purpose it serves, and it happens when you activate, re-validate at launch, or deactivate. If you never enter a key, this never happens. If you are offline, WhisperType fails open rather than locking you out.
Sending a report. Only when you explicitly tap Send — see below.
That is the complete list. There is no analytics SDK, no crash reporter phoning home, no update ping and no advertising identifier.
If you report a bug
WhisperType includes an optional, user-initiated bug reporter, which also sends feature requests. Nothing is ever sent in the background — only when you tap Send. A basic report contains your message, the app and macOS version, and an email address only if you type one. You can optionally include a diagnostics profile — it is off unless you tick it — with your Mac model and chip, CPU core count, RAM, language and region, time zone, and thermal state; it contains no name, account, files, location, or any persistent identifier that ties two reports together, and the panel shows you the exact contents before you send.
Attachments. You may attach screenshots or a zip (up to 15 MB) to a report. Those files are uploaded to our private report storage when you press Send, and only then — they are never gathered automatically, and WhisperType never takes a screenshot by itself. Only attach what you are willing to share: if a screenshot happens to show a document, an email or a transcription, that content goes with it. Remove an attachment before sending and it is never uploaded.
The permissions WhisperType asks for
WhisperType asks for Microphone access so it can hear your dictation, and Accessibility access so it can paste the text into the focused field and let esc cancel from other apps. For the live preview it also asks for Speech Recognition, which is what lets macOS's on-device recogniser run; decline it and everything else works as before. All three are macOS platform permissions you grant in System Settings, and both are local — they let WhisperType do its job on your Mac and send nothing anywhere.
Children's privacy
WhisperType is not directed at children and does not knowingly collect information from anyone, of any age — there's simply nothing collected, full stop.
Changes to this policy
If this policy ever changes, the update will be posted on this page with a new "Last updated" date.
Contact
Questions about this policy? Email contact@hanncrest.com.