Voice to Screen lets you speak on your iPhone and insert the text into a Mac you paired. We designed it to collect as little as possible.
What stays on your devices
Recordings and drafts are stored on your iPhone, encrypted by iOS. Audio is deleted 24 hours after a successful transcription by default; audio whose transcription failed is kept for 30 days so you can retry. You can delete any draft and its audio at any time.
With On this iPhone processing, audio is transcribed on the iPhone and never leaves it.
The Mac companion keeps a small local journal of insertion results (not the text) so it never inserts the same dictation twice.
Dictation using the Mac's own microphone uses Cloud processing through our service and Microsoft Azure. Audio is held in memory on the Mac during dictation and discarded afterwards. The Mac requires internet and a paired iPhone with an active Pro subscription; the iPhone can be offline. Streaming can be turned off, and filler-word removal enabled, in Mac Settings → Dictation on this Mac.
What passes through our service
Cloud processing: audio is sent through our Cloudflare Worker directly to Microsoft Azure Speech (MAI-Transcribe-2 or MAI-Transcribe-2-Streaming), then the text is returned. With streaming enabled, audio is sent while you speak; with streaming disabled or transcription options that require a completed recording, it is sent after you stop. Streaming is on by default and can be turned off in Settings → Processing. Cloudflare and Microsoft process audio and text to handle the transcription request under their respective policies. Our service does not store audio or transcripts, and we do not log request bodies. Cloud audio is not end-to-end encrypted.
Delivery to your Mac: the final text is end-to-end encrypted on your iPhone with keys that only your paired Mac holds. Our relay stores only ciphertext, for at most 24 hours or until the Mac answers.
Readiness: your Mac sends its name, whether it is ready, and the name of the frontmost app to your paired iPhones, end-to-end encrypted. Field contents are never sent.
What we store
Device public keys, device names, which devices are paired, last-seen times, and revocations.
Usage counters (seconds of dictation) for the free trial, and your App Store subscription status (transaction id, product, expiry). We never receive your payment details.
If you use invites: your iPhone's invite code, and which two devices an invite linked, so each invite adds minutes once. We never show who owns a code.
Short-lived security records such as request nonces and rate-limit counters.
On iPhone we may use Apple DeviceCheck to remember, on Apple's servers, that the free trial was used on that device, and that it already used an invite code. It does not identify you.
Feedback
When you send feedback from the app, we email your message to the developer with the app version, operating system version and device model, plus your email address if you choose to add one. The service does not keep a copy.
What we never collect
No accounts, contacts, location, advertising identifiers, analytics of your content, or screen contents. The app asks for the camera only to scan a pairing code, the microphone only while you dictate, and on the Mac only Accessibility to insert text into the field you are using.
Deleting your data
In Settings → Privacy, Delete my cloud account data revokes the device, removes its pairings, queued ciphertext, usage counters, invite records and subscription links. We keep a revoked-device tombstone (device id and public keys, no name) so the identity cannot be reused, and the DeviceCheck trial and invite bits if they were set. Subscriptions are managed and cancelled in your Apple ID settings.