Tunnel.ai

Description (4000 chars max)

Tunnel.ai turns your train of thought into text, wherever you are, even when the train is in a tunnel.

It is built to catch a thought the moment it lands, with as little friction as possible. Tap once and talk. Everything runs on your device, so there is no account, no cloud, and no internet needed after the first launch. Your audio never leaves your phone or Mac.

CAPTURE WITHOUT BREAKING YOUR FLOW • Record over your music. Start a note without pausing your music, podcast, or call. Tunnel.ai captures your voice on top of whatever you are already listening to, so you can talk through a follow-up while the track keeps playing, then get on with your day. • Keep recording when you switch apps. Jump to your notes or calendar to check a name mid-thought and the recording keeps rolling, so nothing gets cut off. • Fully offline. Planes, trains, basements, dead zones. The model lives on your device, so a signal is never required.

WHY ON-DEVICE MATTERS • Private by architecture. There is no server, so we could not read your notes if we wanted to. • No subscription. Nothing to renew, because there are no cloud costs to cover.

WHAT IT DOES • Tap once, talk, tap again. Your words appear as a note. • Confidence highlighting. Words the model was unsure about show in orange. Tap one to correct it. • Custom dictionary and vocabulary. Teach it the names and jargon you actually use, once, so it hears colleagues, brands, and technical terms right. • AI Cleanup on supported devices. An optional on-device pass that fixes misheard names using your own word list. Still nothing leaves the device. • Audio kept with every note. Replay a recording, or re-transcribe it any time, for example after improving your dictionary. • Edit, copy, share, search. Transcripts are plain text, ready for wherever they need to go.

Powered by open-source speech recognition (Whisper, via WhisperKit), accelerated by Apple silicon.

Requires iPhone with iOS 17 or later. [iPhone listing] Requires a Mac with Apple silicon (M1 or later) and macOS 14 or later. [Mac listing] The transcription model is built in, so it works from the first launch with no download.