Updates & release notes

Earshot has captioned the same 99 languages since day one. What’s been uneven is how well — languages with less training data behind them have always trailed the majors. Thanks to feedback from users around the world, Earshot 1.0.2 takes its biggest step yet toward closing that gap.

Earshot now supports Whisper Large Turbo. In Settings → Transcription model, under “Max accuracy,” you’ll find Whisper Large Turbo — OpenAI’s Whisper large-v3-turbo, running entirely on your iPhone. We added support for it for one reason: better captions in lower-resource languages. It’s the largest and most accurate model Earshot offers, and the gains are biggest exactly where they’re needed most — English improves modestly, while Swedish, Swahili, Welsh, Yiddish, Māori, and dozens of others see the largest improvements.

It’s a 626 MB one-time download and needs a recent iPhone (6 GB or more of memory); the first load takes a few minutes while iOS optimizes it for your device. Base and Small remain great everyday choices, and you can switch between downloaded models instantly.

Pro tip: pin your language. By default, Earshot auto-detects the language being spoken. For lower-resource languages you’ll get noticeably better results by turning auto-detection off — Settings → Language → choose your language. Switch back to Auto-detect when you’re traveling or in mixed-language company, and Earshot will follow whatever it hears, in any of the 99.

Accuracy still varies by language, accent, and listening conditions — but with this update, the floor just came up a lot.

As always: free, private, and entirely on-device. Your conversations never leave your phone.

Download Earshot on the App Store →

As of today, Earshot is available on the App Store in 175 countries and regions — including the entire European Union, which required an extra round of Digital Services Act trader verification.

That’s the full worldwide footprint: free, private, on-device live captions in 99 languages, wherever the App Store operates.

Download Earshot on the App Store →

Earshot is on the App Store. Download it free.

Earshot captions speech and flags ambient sounds — a crying baby, a smoke alarm, a doorbell — in real time, in 99 languages. It’s free, fully private (nothing ever goes to the cloud), and works with no Wi‑Fi or cellular connection.

It took 32 builds, three App Review rejections, and a lot of generous feedback from people who tried it along the way. Thank you — truly.

Getting real-time speaker separation (diarization) right was harder than I expected, and honestly it’s still not 100% there. It’ll keep getting better.

Earshot stands on open source. With deep gratitude to:

• OpenAI, for Whisper — the on‑device AI speech model under the hood.
• Argmax, for WhisperKit — the open‑source, iOS‑optimized Core ML harness that makes Whisper fly on iPhone.
• pyannote, whose speaker‑diarization models power multi‑party speaker separation.

Download Earshot on the App Store →

While I await App Store approval, I’ve finally got multiparty speaker separation (diarization) functioning in a reasonable state for 3-person conversations. I tested it on a YouTube video (ideal conditions) — it lags a little and is by no means perfect. It doesn’t pick up the turns as instantly as I’d like — but — it does pick them up. I’m using pyannote, an open-source diarizer, to separate up to four speakers.

My earlier attempt this week failed after I bought an “influencer” 4-mic lavalier-style mic that sadly combines the channels into one, so it was of no use. But now I have a set of mics should I ever become an influencer. 😂

I’m really close to getting Apple’s approval — had to clean up some minor metadata issues. Next steps will be working on UI fit and finish, and seeing how this thing takes off. I’ll keep the core functionality free, and maybe keep the non-core functionality free too 😂 — not sure. Let’s see where this thing goes.

DM or email me if you want early access via Apple’s TestFlight. I’d love feedback.

Earshot has moved to its own domain: tryearshot.ai. Same app, same promise — free, private, and fully on-device — now with a proper home of its own.

This page is where I’ll post what’s new as the app evolves: new features, accuracy improvements, and the road toward the App Store. Earshot is currently in TestFlight beta while I polish the last details.

Want in early, or have feedback? Get in touch.