It listens to the tab’s own audio, so it works where caption translators can’t — Twitch, YouTube Live, lectures, your own files.
Real speech-to-text
Every other subtitle translator needs a caption track to read. Twitch doesn’t have one, and most VODs don’t either.
No caption track to read
Nothing to translate means nothing on screen. On Twitch, that is every stream.
次のマップ、ちょっと怖いかも…みんな一緒に行こう?
The next map might be a little scary… come with me, everyone?
It transcribes the tab’s audio, then translates that. If the tab makes sound, it can be subtitled.
How it works
Install it, sign in once, and press start on whatever is playing. Nothing else to configure.
No API keys, no card. Thirty free minutes are waiting.
このボス、まじで強すぎるんだけど…
This boss is seriously way too strong…
It listens to the audio itself. No caption track needed, ever.
I’m reading all the chat messages — thank you!
Drag it, restyle it, and save the transcript as TXT, SRT or VTT.
Built for watching
Everything here exists because a subtitle that arrives late, in the wrong place, or with the wrong name in it is worse than none.
Drag it anywhere on the video, resize it, change the font and size, toggle the original line. Only the latest lines stay on screen.
It waits for a natural pause, so lines read as whole thoughts. Quick, Balanced and Precise presets.
A glossary locks the spelling of streamers, characters and recurring terms — for the whole session.
The spoken language is detected for you, so there is usually nothing to set.
Drop in an MP4, MP3 or WAV: watch it with live subtitles, or turn it into an SRT or VTT you keep. The file never leaves your computer.
Sync mode
Every live translator is late: it has to hear a line before it can translate it. Sync mode pre-processes the audio instead, so the line is ready before it reaches you.
A translator that leaves playback untouched can’t translate a line it hasn’t heard, so waiting less means choppier lines and waiting more means later ones — a documented trade-off in simultaneous-translation research. Live Sync steps off that curve by holding playback back a few seconds instead.
Short, quick phrasing so each line lands as the words are spoken. Pick it once and you are done.
Waits a beat longer so lines read as fuller sentences instead of fragments.
Picture and sound stay together, so the subtitle never runs ahead of the face on screen.
Sync presets play a touch behind live and do more work on your machine, so they use more CPU and battery than the standard ones — and they cannot run on DRM-protected sites like Netflix. Prefer to leave playback completely untouched? Quick, Balanced, and Precise work exactly as before.
Under the hood
Step one — hearing it right
Every subtitle starts as a transcript, so this sets the ceiling for everything after it.
Published benchmarks, August 2026. Lower is better. Every engine scores worse on a noisy live stream than on a benchmark clip.
Step two — saying it well
Read a line that arrives half-finished
Spell a name the same way an hour later
Answer in under a second, a dozen times a minute
Pricing
Thirty free minutes a week, plan or no plan. Want more? A monthly plan brings minutes every month — first month half price, cancel anytime.
Sign in with Google. Enough for a stream, a lecture, or a few clips a week — topped back up every Monday.
Start freeMinutes land every month, no meter to watch. First month $2.99. Cancel anytime — the month you paid for runs to its end.
Get more minutesTop-ups for plan members and existing buyers: up to 1,440 minutes for $24.99, and they never expire.
Get more minutesConnect OpenAI, Anthropic or ElevenLabs keys and pay only your provider. Audio goes straight from your browser to them.
The Sync presets are a Premium feature — your free weekly minutes include them, and any payment, a first month on a plan included, keeps them on your account for good.
Privacy
No analytics, no browsing history, no stored audio. Payments happen on PayPal’s or Stripe’s own pages. With your own keys, nothing ever touches our servers.
The privacy policy and terms of service are available in English only.
FAQ
Still stuck? The support form is in the footer and goes straight to the developer.
Thirty minutes of live translated subtitles every week, refilled each Monday, with no card and no AI accounts. Past that, a monthly plan brings 600 minutes for $5.99 or 1,500 for $12.99, half price for the first month and cancellable at any time. Or bring your own API keys and pay only your provider.
No. Captions appear only if the streamer runs captioning software, which most don’t, and there’s no translation for viewers at all. Friskription works on your side of the screen, so it doesn’t depend on the streamer.
Yes — that’s the whole point. It transcribes the actual audio of the tab, so no caption track is ever needed. Twitch, YouTube Live, Kick, uncaptioned VODs, lectures: if the tab plays sound, it can be subtitled.
That’s the most popular use, along with Korean streams. The glossary keeps streamer and character names spelled the way the community spells them, and speaker labels keep collabs readable.
47, including English, Japanese, Chinese (Simplified and Traditional), Korean, Spanish, French, German, Arabic, Hindi and Polish — the full list is in the extension. The spoken language is detected for you, so there is usually nothing to set.
Live Caption covers a limited set of source languages, renders into a small fixed box, and translates with basic machine translation. This uses best-in-class speech recognition plus AI translation that reads context, keeps names consistent, and draws subtitles as an overlay you can move and restyle.
No. Audio is processed only while translation is running and is never stored. With your own keys it goes straight from your browser to your provider and never touches our servers. See the privacy policy.
Thirty free minutes, no card, no AI accounts. If it doesn’t help, you’ve lost nothing.
No analytics, no browsing history, no stored audio.