So I'm not sure about the rest of the world, but the United States has something called the "Americans with Disabilities Act (ADA)", which enshrines a bylaw requiring all those services to have closed captions for deaf/hard of hearing viewers. What some companies have attempted to do (Netflix comes to mind, but it's not just them, it's everyone) is if they have multilingual programming, pretend that the English subtitles can do double duty with the English closed captions.
SUBTITLES =/= CC
Subtitles are to match the translation of the foreign language voice over. CC are to match the English dub voice over. If you ever encounter such a discrepancy, report it. This sort of corner cutting is illegal.
What saddens me about Youtube specifically is that not only is their CC service raw MT--with questionable success depending on the voice over--it sometimes overwrites the efforts of creators who went out of their way to make subtitle options in the first place. Even if their subtitles weren't the best, they bent over backwards to try. Now the only ones where I'm not asking myself if it's Google Translate's engine working the captions are the video makers who hard code their subtitles directly into the video files.
Slator Language Industry News did another article many years ago about how Google Translate and other equivalent services that they're no longer able to distinguish between their own MT output and actual human created content when the web crawlers come back with more input data. Google Translate has been Ouroboros -ing itself since at least 2016.
In terms of MT more broadly, it's everywhere. It's been everywhere since the early 2010's if not earlier. It just wasn't nearly as cheap to do at scale back then with the accuracy required to not waste everyone's time. If you go on a foreign language website and your browser asks you if you want to read it in English, that's an MT engine. (If it's a pop up, then often times it's the company asking if you want to be redirected to the segment of the website designed for your IP address' locale.) If you can think of any major company or brand that has multiple language options nowadays, odds are, the entire website was not translated manually. It was translated by MT and then someone edited it after the fact so the output sounded slightly less janky. Companies who have brand specific requirement might do multiple editing passes after that, and change some of the terminology for search engine optimization, if the machine wasn't provided those parameters initially.
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
timeasmymeasure
Re: subtitles
Date: 2025-05-10 08:05 pm (UTC)SUBTITLES =/= CC
Subtitles are to match the translation of the foreign language voice over.
CC are to match the English dub voice over.
If you ever encounter such a discrepancy, report it. This sort of corner cutting is illegal.
What saddens me about Youtube specifically is that not only is their CC service raw MT--with questionable success depending on the voice over--it sometimes overwrites the efforts of creators who went out of their way to make subtitle options in the first place. Even if their subtitles weren't the best, they bent over backwards to try. Now the only ones where I'm not asking myself if it's Google Translate's engine working the captions are the video makers who hard code their subtitles directly into the video files.
Slator Language Industry News did another article many years ago about how Google Translate and other equivalent services that they're no longer able to distinguish between their own MT output and actual human created content when the web crawlers come back with more input data. Google Translate has been Ouroboros
-ing itself since at least 2016.
In terms of MT more broadly, it's everywhere. It's been everywhere since the early 2010's if not earlier. It just wasn't nearly as cheap to do at scale back then with the accuracy required to not waste everyone's time. If you go on a foreign language website and your browser asks you if you want to read it in English, that's an MT engine. (If it's a pop up, then often times it's the company asking if you want to be redirected to the segment of the website designed for your IP address' locale.)
If you can think of any major company or brand that has multiple language options nowadays, odds are, the entire website was not translated manually. It was translated by MT and then someone edited it after the fact so the output sounded slightly less janky. Companies who have brand specific requirement might do multiple editing passes after that, and change some of the terminology for search engine optimization, if the machine wasn't provided those parameters initially.