For the AV desk

Translate what is already going through the desk

If your service or conference already streams, the audio is mixed and ready. Point your encoder at a second destination and the translation runs off the same feed — no laptop in the hall, no browser tab anybody has to keep awake.

Start translating freeSee how it connects
Translation activeExample
Server
rtmp://…
Stream key
••••••••••••••••
Why it exists

A browser tab dies. An OBS or vMix feed does not.

In browser mode the audio comes from a microphone the console captured, so closing the tab or reloading the page ends the capture — the broadcast pauses and waits for somebody to come back. In stream mode the source is an RTMP ingress that belongs to the broadcast itself, so the console is just a window onto it: close it, reload it, take it home, and the translation keeps running on the encoder's audio. That is the whole reason the mode exists, and it is why a venue with a real sound desk should use it.

  • Works with the console closed
  • A fresh stream key for every session
  • Audio only — no video is transcoded
  • The same translation path as browser mode

Connecting it

  1. 1

    Switch the organization to stream input

    One setting in the console, changed before you go live — it cannot be switched during a broadcast, because the source of a running session is not something to swap underneath it.

  2. 2

    Copy the server URL and the key

    Both appear in the console when the session starts. In OBS they go into Settings → Stream as a custom server; in vMix they go into the streaming destination. It is an ordinary RTMP target, so anything that can stream can send to it.

  3. 3

    Start the encoder, then check one language

    Open the listener page on a phone and pick a language. If you hear the room a few seconds behind, the path is correct end to end and you can close the console.

  4. 4

    Add a second destination, not a second stream

    Most encoders push to several targets at once, so your usual stream keeps going untouched. If yours does not, a second encoder instance on the same mix works too.

At a glance

Protocol
RTMP
Stream key
One per session
Video sent on
None
Survives a reload
Yes

What the desk gets

Everything here follows from the audio arriving over the network instead of out of a browser tab.

The mix you already trust

Whatever leaves the desk is what gets translated: gated, compressed, without the room, without the person coughing next to a laptop microphone.

Nothing to keep awake

No tab, no screen saver, no volunteer who must not close the wrong window. The broadcast belongs to the session, not to a browser.

Audio only, on purpose

Your RTMP stream may carry video; we never subscribe to it. Nothing is transcoded on our side and no picture reaches the model or the listeners.

A key that expires with the service

The ingress and its key are created for one session and torn down when it ends, so a key written on a sticky note stops working by itself.

An interpreter can still join

Stream mode changes where the source audio comes from, not what happens after. A person can still take one output language, from anywhere, on their own device.

The same numbers as before

Audience per language, the listener curve, transcript and optional recording all work identically. Only the input changed.

Common questions

Which encoders work?

Anything that can push RTMP to a custom server: OBS, vMix, Wirecast, a hardware encoder, or the streaming output of a video mixer. There is nothing AnyLang-specific to install — you are pointing an ordinary RTMP destination at a URL and key we give you.

Do you receive our video?

Your stream may contain video and we never subscribe to it. Only the audio track is read, so no picture is transcoded on our side, none reaches the translation model, and none is delivered to listeners. If you would rather not send it at all, configure an audio-only output in your encoder — either way the result is the same for us.

What happens if the encoder drops?

The broadcast stays up and the listeners keep their connection; they hear silence, exactly as they would during a long pause. When the encoder reconnects to the same key, audio resumes. The session is not ended by a dropped input — only Stop ends a session.

Can I keep the same stream key between services?

No. The ingress and its key are created for one session and torn down at the end of it, which is what stops an old key on a sticky note from reaching a broadcast it has nothing to do with. Copy the new key from the console at the start of each session.

How much delay does this add?

Very little of its own — the RTMP hop is small next to the deliberate pause on the output. The translated voice is held for a few seconds so a whole clause arrives at once rather than word by word, which is what makes a long service followable. Captions in the language being spoken are not held back at all.

Should everyone use stream mode?

No. If you have no encoder and nobody who configures one, browser mode is simpler and sounds fine: pick the microphone in the console and start. Stream mode is worth it when there is already a desk, already a stream, and somebody who would rather not depend on a laptop staying open at the back of the hall.

Read next

Point your encoder at it

30 free credits, no card. Enough to test the whole path end to end before a real service.

Start translating free