Viewers can caption or translate any stream they're watching; streamers can show captions to everyone. Two setups, one tool, moments behind the voice.
Uploaded videos get captions eventually; live streams mostly go out raw. Twitch has no built-in captions, and a YouTube stream's automatic captions — where they appear — arrive on the platform's terms, in the stream's language. Meanwhile the audience is more international than the audio: viewers watching an English streamer from São Paulo, watching muted at work, watching through hearing loss. Captioning a live stream is two different problems — the viewer captioning what they watch, and the streamer captioning what they broadcast — and both have a clean setup. Pick your side; this guide covers each in turn.
Everything about the setup follows from one question: whose screen should the captions appear on?
They're independent — a viewer can caption a stream that already shows captions, for instance to translate it — but the setups differ, so they get separate walkthroughs.
YouTube, Twitch, or anywhere else — for this setup a stream is just a tab playing audio. Leave it playing at normal volume.
In Unicaption, choose tab or system audio as the source. It reads the stream's sound digitally — your room, your speakers, and your mute button don't affect it.
Watching a streamer in a language you half-know? Set your target language and read the original and translation side by side, from 60+ languages, with auto-detect handling the rest.
Stream in one window, captions beside it. They run moments behind the voice — close enough to feel attached to the speech.
Because capture is digital, this setup works with the stream muted — captions keep flowing while the sound stays off, which is quietly the killer feature for watching at work, next to a sleeping kid, or in any room the stream doesn't belong in.
Streamers add captions the same way they add any visual element: run a caption view on the streaming machine, then capture that window as a source in your streaming software — OBS or anything else that can capture a window or region — and place it in the scene. The captioning tool listens to your microphone; the broadcast shows the result to everyone.
A stream's audience is self-selected from the whole planet, so the mismatch between one streamer's voice and everyone's ears is at its widest here. Viewers follow streamers across language lines for the game, the craft, or the personality — comprehension trailing loyalty. Accented speech that's easy for some viewers is work for others, and a transcript resolves what the ear can't: unfamiliar pronunciation becomes familiar spelling. Deaf and hard-of-hearing viewers get a live medium that has mostly excluded them; muted viewers get the stream anyway.
For streamers, captions and translation are audience infrastructure — the difference between a stream your international viewers tolerate and one they can actually follow. The same attendee-side-scales logic we describe for webinars applies to streams: you broadcast captions in one language, and each viewer who needs a different one runs their own translation.
Unicaption covers both directions with one switch — the audio source:
It runs in the browser — a viewer needs no install at all — and the free tier gives you 30 minutes every week with no credit card: one evening's stream is enough to judge it. Setup basics live in our getting started guide.
Streams are acoustically hostile: music beds, game audio, hype moments where everything peaks at once. Viewers captioning from tab audio inherit whatever mix the streamer made, so if the music drowns the voice for your ears, it drowns it for the engine too — captions will favor whatever is loudest. Streamers control the fix: keep your voice several dB proud of the music, duck the soundtrack when you talk, and treat caption quality as a mix decision. The deeper tuning checklist — vocabulary, levels, mic habits — is in our caption accuracy guide.
Twitch has no built-in captions for most streams, but you can caption any stream yourself: open it in a browser tab, run Unicaption alongside it, and choose tab or system audio as the source. Captions appear on your screen moments behind the streamer, and you can add live translation into your language from 100+ options. The streamer doesn't need to do — or know — anything.
Yes. Open the stream in a tab, point Unicaption at that tab's audio, and set your target language: the original transcription and your translation render side by side as the stream plays, with automatic language detection. This works on any live stream your browser can play, regardless of what caption features the platform itself offers.
Run a caption view on your streaming machine, listening to your microphone, then capture that caption window as a source in your streaming software and place it in your scene — the same way you'd add any overlay. Caption your mic rather than the full mix so game audio and music stay out of the transcript, and add game and community vocabulary to Unicaption's custom dictionary before going live.
Yes — this is one of the best reasons to use them. Tab and system audio capture read the stream's audio digitally, before your volume controls, so captions keep flowing with the sound off entirely. You can follow a stream in a quiet office or next to a sleeping kid with the audio muted the whole time.
Caption what you watch or what you broadcast — live, translated, in 60+ languages. 30 free minutes every week, no credit card.
Start Free Trial →