You
Excuse me — do you know where the nearest convenience store is?
すみません、一番近いコンビニはどこかご存じですか？
How it works
Two earbuds.
One conversation.
metcha is a live translator that runs between a single pair of earbuds. Each person keeps one — and each hears the other in their own language, in their own ear.
The three-step flow
- 1
Share an earbud.
Keep one for yourself. Give the other to the person you want to speak with. AirPods, Beats, Sony, Bose, Samsung Galaxy Buds — anything that pairs with your iPhone.
- 2
Pick two languages.
Choose what you'll speak and what they'll speak. metcha handles the rest — language detection, transcription, translation, and voice synthesis.
- 3
Talk normally.
Speak in your own language. You hear them in yours; they hear you in theirs. The phone goes in your pocket and the conversation carries on.
What it sounds like
A snippet from an actual metcha session — English ↔ Japanese, asking for directions in Tokyo. Each speaker hears the other's line in their own language.
Them
駅の右側に出てすぐ、ローソンがあります。
Right outside the station on the right, there's a Lawson.
You
Perfect, thank you very much.
完璧です、ありがとうございます。
What's happening under the hood
metcha runs four stages, end to end, every turn:
- Capture. Your iPhone's microphone picks up your voice. metcha uses Apple's built-in voice-activity detection so it knows when a sentence ends without you tapping anything.
- Transcribe. On the free tier, Apple's on-device Speech framework converts the audio to text. On metcha Plus, Deepgram's cloud STT handles the same step with higher accuracy across accents and noisy environments.
- Translate. Apple's on-device Translation framework handles the free path. metcha Plus optionally routes through a Claude-powered translation for more natural, idiomatic phrasing, with a quiet fallback when the network is slow.
- Speak. The translated text is spoken into the other person's ear — using a built-in iOS voice on the free tier, or an ElevenLabs premium voice (optionally cloned from your own) on metcha Plus.
For the deeper version — latency budget, fallbacks, what stays on-device — see the methodology page.
Common questions
- Do both people need to install metcha?
- No. The simplest mode runs on one phone: you keep one earbud, give your conversation partner the other, and metcha handles both sides. If you each want your own earbud, the second person can join from a browser by scanning a QR — still no install for them.
- Does metcha work without an internet connection?
- Yes, for the language pairs Apple supports on-device. The free tier uses Apple's built-in Translation framework, which downloads language packs to your phone. metcha Plus features that depend on cloud voices or cloud translation need a network connection.
- Which earbuds does it work with?
- Most major Bluetooth earbuds — AirPods, AirPods Pro, Beats, Sony, Bose, Samsung Galaxy Buds, and similar. metcha is hardware-agnostic. If your earbuds pair with your iPhone, they work.
- How much latency is there?
- A short pause — usually under two seconds for the on-device path, slightly longer when cloud translation or premium voices are in the chain. Long enough that you wait a beat, short enough that the conversation still feels like a conversation.
- Is the conversation private?
- metcha does not store your conversations on our servers and we will never use them to train a model. Cloud-assisted features pass audio through carefully chosen third-party services (Deepgram for speech-to-text, ElevenLabs for premium voices, an LLM provider for the optional Better Translation path). See our privacy policy for the full picture.
- What languages are supported?
- metcha supports English, Spanish, French, German, Italian, Japanese, Romanian, and Russian at launch, with more on the way. The exact mix of on-device translation, cloud STT, and premium voices varies by language — see the languages directory for the per-pair breakdown.
Try it on your next conversation.
Free to download. Free to use the on-device path forever.
Download on the App Store