Every voice.
Every sound.

Real-time captions that are truly free, fully private, and run entirely on your iPhone. No accounts. No cloud. No tracking.

Coming soon to the App Store Request early access →

FREE· PRIVATE· OFFLINE· 99 LANGUAGES· ON-DEVICE

languages

300+

sounds recognized

bytes of audio uploaded

100%

on-device

The story

From a product manager behind Android’s Live Transcribe.

In 2019, a team at Google spent nine months building Android’s Live Transcribe — free, real-time captions for speech and ambient sound, inspired by a Deaf researcher who’d spent his career on captioning. It launched at Google I/O, became one of the most-adopted accessibility apps Google ever shipped, and came preloaded on billions of devices. I was one of two product managers on it.

Google never brought it to iOS, and nothing has truly filled the gap — the lookalikes charge, upsell, or send your audio to the cloud. Having a disability shouldn’t cost you money or your privacy.

So I rebuilt it for iPhone. Free. Private. Entirely on-device. That’s Earshot.

Free

No sign-up, no sign-in, no ads, no upsells. Free should mean free.

Private

Everything runs on your device. Your audio is never uploaded, logged, or stored.

Offline

Works with no Wi-Fi or cellular. The speech model lives on your phone.

99 languages

Caption in 99 languages — and translate any of them into English on the fly.

Sound-aware

Recognizes hundreds of ambient sounds, with full-screen alerts for smoke alarms, doorbells, a crying baby, and more.

Built for accessibility

Designed with the deaf and hard-of-hearing community first: Dynamic Type, VoiceOver, high contrast.

Earshot live captions screen — Live captions, big and readable.

Earshot smoke-alarm alert — Full-screen alerts for urgent sounds.

Earshot sound catalog — Hundreds of recognizable sounds.

Your words never
leave your phone.

No accounts. No analytics. No tracking. The only network request Earshot ever makes is a one-time download of the on-device speech model — OpenAI’s Whisper. After that, it works fully offline.

Read the full privacy policy →

Under the hood

Real speech recognition, running on your iPhone.

Earshot transcribes locally with OpenAI’s Whisper. Word Error Rate — the standard accuracy benchmark — is about 8.6% for the small English model. It also ships faster tiny and base models that fly on older iPhones, so you can trade a little accuracy for speed when you need to.

Read along with the world.