Veilr is a random video, voice and text chat with strangers around the world where the language barrier is handled for you. Messages can arrive translated into the reader's language, and on voice and video calls you see live captions of what the other person is saying, translated into yours. It runs in the browser and on Android, it's anonymous by default, adults-only, and free to start.

Why it exists: an Omegle alternative where language doesn't matter

The best conversations on the internet happened between strangers who would never have met otherwise. The original Omegle proved that for fourteen years and shut down in 2023. It left one problem nobody had solved: most of the world still couldn't understand each other. Veilr is built around that one problem.

How it works

Pick a format: text, voice or video. Choose the language you're practising and a few interests, and the matcher ranks people toward you who share them — including native speakers of the language you're learning, a language exchange with captions as a safety net. Then you just talk. In text, you type in your language and your message can reach them in theirs, with the original one tap away. On a call, a streaming speech-to-text pipeline turns what they say into captions translated into your language — 34 caption languages, and it works in a voice-only call too, where there's no camera at all. You hear their voice. You read their words. And when you don't need the captions anymore, you switch them off.

Filters are preferences, not walls. On a free account, gender, age and country preferences hold for a few seconds and then relax if nobody matching is online, so you're never stuck in an empty queue; the other person's preferences are always respected. A Boost pass turns your preferences into guarantees and gives you first pick.

Anonymous chat, adults-only

The person you're matched with never sees your name, email or the account you signed in with. We ask for a date of birth at sign-up and refuse a match to anyone recorded under 18. Anyone can report or block in one tap. A report about a suspected minor skips automated analysis entirely and goes to a person; other reports go through automated analysis and reach a person when they escalate. The text of a random chat is not written to our permanent database — a short buffer of the last messages expires within minutes of the chat ending. The one exception is a choice you both make: if you add each other as contacts, that conversation stays, and you can message and call each other again — still translated, still captioned. That's how a stranger becomes the person you practise Spanish with on Tuesdays.

Where it runs and what it costs

Veilr works in any modern browser with a 16-language interface, and on Android via Google Play (not available in the UK yet; on iPhone it runs in the browser). It's free to start: every account gets a run of message translations and five minutes of live captions. Boost passes lift the ceiling — $1.99 for an hour, $4.99 for a day, $9.99 for a week, $19.99 for a month — and a caption pack adds 120 minutes for $2.49. Sign-in with a social account is required; it's the abuse control a one-person project can afford.

Bring people, get paid

Every account has a referral link. New accounts earn referral rewards as free in-app credit; switch the reward form to money in Invite friends → Affiliate program and you earn 20% of the list price of what your invitees buy, paid in USDT after the 30-day refund window (payouts from 100 USDT).

Under the hood

WebRTC with perfect negotiation (glare rollback, ICE restart, stable m-line order); self-hosted coturn plus Cloudflare's TURN as relay — for some phone-to-phone pairs a relay isn't a fallback but the connection itself. A WebSocket layer that runs multi-pod via Redis pub/sub, so two people on different pods still get their signaling relayed. Per-format Redis queues with a JS matcher, and the final pair claim in one atomic Lua script so two pods can't take the same partner. Speech-to-text by Soniox, text translation by Gemini. Node.js, PostgreSQL and Redis on the backend; React, TypeScript and Vite on the front; a Capacitor build for Android; 16 locales.

Honest limits

Veilr is built by one developer and the user base is small, so outside the busiest window — 16:00–20:00 UTC most days — you may wait for a match. I'm in the queue myself. If you try it, I'd love to know which language pair you tested and whether the captions kept up.