IPHONE
undervoice for iPhone
Open the App Store on iPhone and install undervoice. On first launch, follow the prompts for camera, microphone, and Local Network access.
Requires iOS 18 or laterDownload on the App StoreQUICK START / SUPPORT
Install undervoice on both devices, prepare the speech model on Mac, and scan the pairing code once with iPhone. After that, place the cursor in a text field and hold the button on your phone to speak.
iOS 18+ · macOS 14+ · Paired recognition requires an Apple Silicon Mac01 / GET BOTH APPS
Both apps are required. It is easiest to prepare the Mac first, then open the iPhone app and scan.
IPHONE
Open the App Store on iPhone and install undervoice. On first launch, follow the prompts for camera, microphone, and Local Network access.
Requires iOS 18 or laterDownload on the App StoreMACOS
Download the DMG from this website. Open it, drag undervoice into Applications, then launch it from the Applications folder.
Requires macOS 14 or laterDownload for Mac02 / PREPARE YOUR MAC
Speech recognition runs on your Mac, so the model and text-insertion permission must be ready first.
Launch it from Applications. The Mac app also lives in the menu bar and can keep running after its main window is closed.
Open Settings, choose your dictation language, then confirm and download the language pack. Wait until its status is ready.
Allow Local Network and Accessibility. Accessibility is used only to type into the focused field; microphone access is needed only if you also dictate with the Mac microphone.
Choose Devices in the sidebar. The QR code for your iPhone appears there.
03 / PAIR
For the simplest first pairing, put both devices on the same Wi‑Fi network.
ON YOUR MAC
Keep the QR code visible. If you see “iPhone already paired” instead of a code, reset pairing first.
ON YOUR IPHONE
Use the scanner in first-time setup. If you are already inside the app, open Settings → Mac pairing → Scan Mac QR code. The connection continues automatically after you allow Local Network access.
04 / START TYPING
There is no transcript to copy back from another app.
Click an editable field in Codex, Claude Code, Xcode, Terminal, a browser, or almost any other app.
Hold it about 5–10 cm from your mouth. Close capture means you can speak softly instead of addressing the room.
Keep holding while you speak, then release. Encrypted audio goes to the paired Mac; once recognition finishes, the final text appears at the current cursor.
undervoice will not type into password or secure-input fields. If focus moves, it stops insertion so text does not land in the wrong place.
HOW IT WORKS, SIMPLY
You only need four ideas—no networking or model knowledge required.
iPhone microphone captures nearby speech
Audio is encrypted on iPhone
The paired Mac runs a local speech-recognition model
Text goes into the focused field
This is not a chatbot processing your voice. The speech-recognition model on Mac only turns sound into text. undervoice does not operate an audio relay or cloud transcription service, and recordings are not saved as audio files.
COMMON FIXES
The code is hidden after the Mac remembers an iPhone. Open Devices, reset pairing, and scan the new code.
Allow Local Network access on both devices and start on the same Wi‑Fi. Managed Wi‑Fi can block device-to-device traffic; try iPhone Personal Hotspot, or configure Tailscale for cross-network use.
On Mac, open Settings → Speech models, download the selected language pack, and wait for Ready. Paired recognition requires Apple Silicon.
Give undervoice Accessibility access on Mac and place the cursor in a normal editable field. Return to the app and check permission again after granting it.
Tell us which step stopped and what status each device shows. Please do not send pairing codes, recordings, or private transcript content.