Live translation for Twitch
VoxTranslate for Chrome translates the audio of a Twitch stream as it plays, with subtitles over the player and an optional spoken translation. It reads the stream's sound, not its chat, so it works on any channel without the streamer setting anything up.
How it works
Open the channel and press the VoxTranslate toolbar button on that tab. VoxTranslate captures that tab's audio and renders translated subtitles in an overlay that stays with the player in theatre and fullscreen mode. On tiers that produce audio the translation is spoken while the stream's own audio ducks underneath on a volume control you set, which matters more here than elsewhere — a stream's atmosphere is half the point, and muting it to understand it defeats the purpose. Nothing is required from the streamer, and nothing appears in chat.
What this is actually good for
Following a stream in a language you do not speak, when the content is worth the effort: a tournament caster, a developer stream, an interview, a talk being restreamed. Anything where someone is speaking and you want to know what they are saying.
The subtitles sit over the player and stay there in theatre and fullscreen mode, so you are not choosing between the stream and the translation.
Why the original audio control matters more here
On a webinar you might happily mute the original. On a stream, the original audio is a large part of why you are there — the crowd, the game, the streamer’s delivery and timing.
So the mix is a dial rather than a switch. Set the original wherever you like from 0 to 100%; while the translated voice speaks it drops underneath and comes straight back when it stops. Many people run the original loud and use subtitles only.
Be realistic about accuracy
Streams are the worst case for speech recognition, and pretending otherwise would waste your money. Game audio over voice, two people talking at once, community-specific slang, shouting — all of it degrades transcription before translation even starts, and the translation cannot recover what was never heard correctly.
It works well for someone speaking clearly into a decent microphone. It works badly for a chaotic four-person party stream, and we would rather say so here than have you find out after the fact.
What it does not do
Live speech in a stream is the hardest thing to transcribe: overlapping voices, background music, game audio, shouting and heavy slang all degrade the result, and no amount of tier selection fixes that. It translates speech, not chat, and not text on screen. Because it bills per minute of speech, it fits watching a specific segment rather than leaving a stream on all evening. Chrome only, and the spoken language is detected once at the start.