Add real-time translation to your YouTube Live stream
You can run a YouTube Live stream with live translated captions — speak one language, viewers read
another. The translation engine is the same OpenAI-compatible pipeline used for any live stream
(Whisper speech-to-text → GPT-5.5 translation via APIVAI → captions in OBS); this guide focuses on
the YouTube-specific setup. For the full pipeline and code, see the AI live-stream translation guide.
What you need
- OBS Studio.
- The translation pipeline writing a caption text file (see the live-translation guide).
- A YouTube channel with live streaming enabled (verify your account; first-time enablement can take 24h).
1. OBS scene with captions
Add your camera/screen, then a Text (GDI+) source pointed at the caption file your script writes
(e.g. caption_es.txt), or a Browser source for styled captions. Place it as a lower-third.
2. Get the YouTube stream key
- Open YouTube Studio → Create → Go Live.
- Choose Streaming software.
- Copy the Stream key (and the RTMPS/RTMP server URL).
3. Configure OBS
- Settings → Stream: Service YouTube - RTMPS, paste the stream key.
- Settings → Output: 3000–6000 Kbps is reasonable for talking-head streams; match your upload speed.
4. Go live
Start your translation script (captions update live), click Start Streaming in OBS, then in
YouTube Studio confirm the preview and Go Live. Viewers now see real-time translated captions.
YouTube notes
- YouTube has its own auto-captions, but they don't translate live the way this pipeline does — your OBS caption overlay gives translated text immediately.
- For VOD, the recorded stream keeps the burned-in captions.
Tips
- Keep captions to 1–2 lines; clear them on silence.
- Short chunks (1–3s) → ~1–2s latency. GPT-5.5 via APIVAI keeps long-stream cost low.
FAQ
Can I translate a YouTube Live stream in real time? Yes — run speech-to-text + GPT-5.5
translation (via APIVAI) and overlay the result as an OBS caption source on your YouTube stream.
Does APIVAI stream to YouTube? No — OBS streams to YouTube; APIVAI provides the translation (the
GPT step). Speech-to-text and TTS run locally.
Which model for live translation? GPT-5.5 — fast, fluent, cheap per call.
Can I stream to YouTube and Facebook at once? Yes — OBS multi-output or a restream service; the
same caption source works for both.
Get started
Set up the translation pipeline (live-translation guide), add the caption source to OBS, and stream
to YouTube Live. Examples: APIVAI examples repo.
