Blog

“Works on Twitch” Doesn’t Tell You Whose Stream It Works On

November 4, 2026  ·  Tablingo

Three kinds of tool all print the same three words. One needs the broadcaster to switch something on, one needs a program on your computer, and one needs neither. The difference decides whether you can use it tonight.

You are watching a Japanese streamer play a game you know well. He laughs at something, the chat scrolls past faster than you can read it, and you understand none of it.

So you search for a way to follow along. Every result claims Twitch support. Three of them are telling the truth, and they mean three different things by it.

Those three words describe where the audio gets picked up. That one fact decides everything downstream: whether you can use the tool on a stream you did not start, whether you have to install a program on your computer, and whether it works on the machine you are actually sitting at right now.

One phrase, three architectures

What the claim meansWhat it asks of youWhat it asks of the broadcaster
Broadcaster-enabled extensionPick a languageInstall it, keep it running
Desktop app (system audio)Install, grant audio accessNothing
Browser extension (tab audio)Install one extensionNothing

All three are honest about what they do. Only two of the three can be used on a stranger's stream tonight, and only one of those two stays inside the browser.

The kind that depends on a stranger

Twitch has carried caption extensions for years. They are genuinely useful, and they work in a way worth understanding: the extension is installed on the broadcaster's channel, and only the broadcaster can add it.

Once it is live, viewers get a language picker in the player. Before it is live, viewers get nothing. There is no button on your side that turns it on for someone else's channel, because the extension does not exist in your browser — it exists in the streamer's setup.

That is a reasonable design for its actual purpose. A broadcaster who wants to caption their own show installs once and serves every viewer afterwards. The cost is paid by one person and the benefit is shared.

But it answers a different question than yours. You are not captioning a show. You are trying to understand one that is already running, made by someone who has never heard of you and has no reason to change their encoder settings for a viewer they cannot see. The "support" in this design is the streamer's support. You cannot supply it yourself.

This is the first thing to check on any product page, and it is usually written in plain sight: does it say the broadcaster has to enable something? If yes, the tool works exactly as advertised and still does nothing for your situation.

The kind that puts a program on your machine

The second architecture gives up on the browser entirely. It installs a desktop application that captures whatever your computer is playing — system audio, loopback, whatever the operating system calls it locally.

This has a real advantage, and it is worth stating plainly: it works on everything. The Twitch desktop client, Spotify, Discord, a local video file, a game with voice acting, a call in an app that never opened a browser. If the machine can play it, the program can hear it.

The costs arrive at install time. On Windows this usually means a loopback capture device, sometimes a virtual audio cable that has to be configured before anything works. On macOS it means granting screen-recording permission, because system audio is captured as part of the screen. Either way you are handing an application the ability to hear everything the machine plays, not just the one tab you wanted translated.

There are practical costs too, and they are the ones people hit first. A work laptop may not allow the install. A locked-down machine will not. Switching computers means installing again, on each one. And the phone in your pocket, where a surprising amount of live viewing actually happens, runs an operating system that no desktop app supports.

None of that makes the approach wrong. It makes it a different trade: broader reach, in exchange for a footprint. The breadth is real. So is the footprint, and it is why these tools can cover every app — they are not inside the browser, so they are not limited by it.

The kind that stays in the tab

The third architecture lives inside the browser and never leaves it. A browser extension asks the browser for the audio of one specific tab.

The mechanism matters more than it sounds. Chrome hands the extension an audio stream for that tab, duplicated before it reaches your speakers. The speakers still get their copy. The extension gets an identical one. Nothing is re-recorded through a microphone, nothing is routed through a virtual device, and the quality is the source quality rather than whatever survives the trip through the air.

Three consequences follow, and they are the reasons this shape exists:

The limitation is symmetrical and should be stated as plainly as the advantages: it only works in a browser. The Twitch desktop client, the mobile app, the console, the TV stick — all outside the boundary. If you watch streams on a television, this shape does not help you, and the desktop-app shape might.

That is the whole trade, in one line: tab capture gives up reach to give up footprint. Neither side is a trick. They are two answers to the question of where the audio gets picked up, and the answer determines which streams you can use them on.

Three questions, thirty seconds

You do not need a comparison table from anyone, including this one. Open the product page and look for three things. Whichever you find first tells you the architecture:

  1. A download button for Windows or macOS. That is the desktop route. It will cover every app on that machine and nothing outside it.
  2. A sentence about the broadcaster, the streamer or the channel owner enabling something. That is the broadcaster route. It works only on channels where somebody already agreed.
  3. The word "tab", or "browser tab audio", or a supported-sites list that reads like a list of websites rather than a list of apps. That is the tab route. It works on any stream you can open in a browser, and on nothing you cannot.

Then do the test that no page can fake. Open a live channel belonging to someone you have never spoken to — any channel, in any language, right now — and try to start captions. If it asks you to contact the streamer first, you have your answer about who the feature was built for.

What each platform already gives you

It is worth knowing how much of this the platform handles on its own, because the answer is different for every site and mostly disappointing:

PlatformCaptions built inTranslation built in
TwitchOnly if the encoder sends themNone
YouTube LiveAutomatic, mostly EnglishNeeds an existing track
Kick, Chzzk, Bilibili LiveNoneNone

Read the third column as a whole and the shape of the problem appears. Where captions exist, translation is a second feature that may or may not be switched on. Where captions do not exist, there is nothing to translate. The gap is not evenly distributed — it is total on most of the platforms where foreign-language streamers actually broadcast.

What we do not do

We take the third route, so the honest boundary is ours to state. Tablingo is a browser extension. It captures the audio of the tab you are watching and puts the original line above the translation, live, while the stream keeps playing.

Outside the boundary: desktop clients, mobile apps, consoles, TVs. If the stream is not in a browser tab, we cannot hear it. We also do not touch the chat — the scrolling text beside the video is a separate stream of language, and translating it is a different problem that we have not solved. We translate what is spoken, and that is all.

None of that is a hidden limitation. It is the shape of a tool that asks nothing of the broadcaster and installs nothing outside your browser, and that shape has a boundary by construction.

The point of all this is smaller than a buying decision. "Works on Twitch" is a sentence with the subject missing, and the subject is the part that decides whether it helps you. Once you know whose stream it works on — the broadcaster's, your whole computer's, or the one tab in front of you — the rest of the listing is detail.

If you want to try ours, it's called Tablingo. The general case for live streams is here, how the two capture strategies differ is here, and why the captions arrive a beat late is here.