The fastest way to add real-time translated subtitles to OBS Studio — drop-in browser source, 30+ languages, under 500ms of latency. No plugin install. No GPU required.
Start free trial →Most "OBS translation plugins" (LocalVocal, OBS-Captions-Plugin, Polyglot) run AI models locally on your GPU. That means:
StreamTranslate runs everything in the cloud. You add ONE browser source URL (the same way you add a Twitch alert or chat overlay) and it works. Across every OBS version since 22. Mac, Windows, Linux. Zero GPU. Sub-500ms latency.
The browser source is transparent — captions overlay on top of your gameplay, webcam, or scene without changing anything.
StreamTranslate listens to your mic input (browser-captured), sends audio to Deepgram Nova-3 for real-time speech-to-text, then races translations across DeepL, Google Translate, Azure, and Lecto in parallel and picks whichever returns first. The translated text gets pushed to your OBS browser source via Socket.IO in under 500ms end-to-end.
You stay in full control of style: font, size, color, outline, position (top/bottom), translation-only mode (hide spoken), live typing on/off. Update settings from the control panel — they sync to OBS instantly.
No. StreamTranslate uses an OBS browser source — no plugin install, no DLL files, no compatibility issues with OBS updates. Works on OBS Studio 22 and above (2018+).
Yes. StreamTranslate has a free 6-hour trial with no credit card required. After that, plans start at $5/month for 50 hours.
Yes. Any streaming software that supports Browser Sources (OBS Studio, Streamlabs Desktop, Meld Studio, Lightstream, XSplit, vMix) works the same way.
LocalVocal runs Whisper locally on your GPU, eating 5-20% of available VRAM during streams. StreamTranslate runs cloud-side — zero GPU impact. Also no plugin install, no OBS version conflicts.
Source language detection in 50+ languages. Target translation in 30+ including Spanish, Portuguese, French, German, Japanese, Korean, Mandarin Chinese, Arabic, Hindi, Russian, Italian, Polish, Vietnamese, Indonesian, Turkish, Thai.
Under 500ms end-to-end. Deepgram nova-3 transcription runs in ~300ms, translation race adds 200-500ms, Socket.IO push to overlay is <50ms. Total: 600ms-1.5s typical, hard cap of 2s.
Yes. The overlay supports showing the original spoken language on top and the translated language below, or just the translation. Toggle in settings.