undervoice: Speech to Text

undervoice is a quiet speech-to-text and voice-typing app for Mac. Speak softly into your iPhone; what you say in English is transcribed locally on the Mac and inserted at the active cursor in Codex, Claude Code, Xcode, Terminal, a browser, or another text field.

It is built for developers and knowledge workers who want the speed of dictation without giving the whole office a performance. Explain a bug, talk through an edge case, or capture a half-formed idea in an AI prompt while it is still fresh—even at a shared desk.

WHY USE IPHONE?

Your Mac’s built-in microphone may be roughly 20–30 inches from your mouth. Your iPhone can be just 2–4 inches away. That closer starting point gives speech recognition a cleaner signal before keyboard noise, ventilation, and nearby conversations compete with your voice. There is no external microphone to buy, wait for, or set up.

HOW IT WORKS

1. Install undervoice on iPhone and Mac.
2. Download the English recognition pack on your Apple Silicon Mac and pair once.
3. Put the cursor where you want the text, hold to talk on iPhone, then release.

OPTIMIZED FOR ENGLISH

undervoice uses a dedicated English recognition path on Mac. The model, local runtime, and product settings are selected for natural English dictation—from quick instructions to long, detailed coding prompts. English is the language being optimized here, not one item in a generic multilingual feature list.

DESIGNED FOR QUIET WORK

• Close-range capture for low-volume speech
• Local recognition on your paired Mac
• A dedicated English recognition path for natural speech and long-form dictation
• Direct insertion at the active cursor—no copy and paste
• End-to-end encrypted audio, transcripts, and control messages
• Audio processed in memory, never saved as recording files
• No undervoice speech cloud
• Screen Protection uses Dark Mode, lowers brightness to 30%, and softens the recording button while the app is active
• A minimal interface that stays out of your coding flow
• Optional local transcript history; nothing syncs to an undervoice server

PRIVATE BY DESIGN

On a local network, audio goes directly from iPhone to the paired Mac, where recognition runs locally. No audio file is left on either device. Password fields and Secure Keyboard Entry are excluded. If focus changes during dictation, insertion stops.

REQUIREMENTS

Requires iOS 18 or later and macOS 14 or later. Paired iPhone recognition requires an Apple Silicon Mac. Both apps are required. Get the Mac companion and setup guide at undervoice.net/en/download.

undervoice is not a chatbot and does not rewrite your ideas. It simply gives your voice a shorter, quieter path into the tools you already use.

TERMS: https://www.apple.com/legal/internet-services/itunes/dev/stdeula/