101
Crunchyroll accidentally confirmed it uses ChatGPT for subtitles (www.pocket-lint.com)
submitted 1 year ago by to c/technology@beehaw.org
30 comments fedilink hide all child comments

all 30 comments

sorted by: hot top controversial new old
[–] 16 points 1 year ago (17 children)

How are subtitles created usually? Are they provided by the source material team, some professional third party that manually transcribes the video, or just fans doing it for free?

  • permalink
  • fedilink
  • source
  • hideshow 17 child comments
    •
  • [–] 47 points 1 year ago* (last edited 1 year ago) (10 children)

    See that’s the kicker, for the longest time it was basically all fan translated subtitles, and only recently have payed for translation become the norm.

    So it’s really quite pathetic for them to try and save a few bucks by replacing a proper translator with a LLM, given that there are still plenty of passionate fans who would have done it for free. Especially given that translating between Japanese and English in a cultural context heavy situation is something these LLMs are really bad at.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 10 child comments
    •
  • [–] 7 points 1 year ago (9 children)

    given that there are still plenty of passionate fans who would have done it for free

    I'd imagine this is a non-starter from a corporate standpoint. I know if I were in charge I'd be terrified of the idea of just trusting community-submitted subtitles to not have random slurs or something inserted. That said I still think it would be super cool if they'd let people source and use their own subtitle files; I now it's possible because I have a tampermonkey script that lets me do just that.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 9 child comments
    •
  • [–] 5 points 1 year ago (8 children)

    That’s the core of the issue, crunchy roll has sat its self as a corporate middleman, buying the rights to distribute shows and then charging consumers a subscription for access.

    But they can’t be bothered to do the only actual damn work their position would realistically demand, beyond renting server space; providing translations for the foreign media they’re distributing.

    That’s without even discussing the fact that not a single penny users give them will end up in the hands of any of the exploited artists who actually made the shows, since the industry doesn’t work on residuals or any other kind of profit sharing, the licensing fees crunchy roll pays essentially going straight to financiers.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 8 child comments
    •
  • [–] 1 point 1 year ago (6 children)

    That’s without even discussing the fact that not a single penny users give them will end up in the hands of any of the exploited artists who actually made the shows

    That's quite the assertion. How exactly do you suggest they're buying the rights to distribute the shows then?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 1 point 1 year ago (4 children)

    They’re buying them from production committees and other such organizations. Most anime is made on essentially “commission” basis, where a studio is payed a fixed upfront amount by a group of financiers and other interests, who then distribute the show, sell the merch, and license it internationally. Essentially studios and those who work there are payed no residuals or other profit sharing scheme like is common in the American film and television industry.

    There is actually a bit of a cartel in that regard, with the third parties that purchase shows from studios having collaborated to suppress the cost of seasons for nearly 2 decades, leading to stagnant wages and rampant overworking of artists as the quality and quantity of work expected increases while the budget stays the same. Increasingly artists at the companies have had to fall back on gig work beyond their standard hours to make ends meet, getting payed by frame in their off hours to make a little extra money, effectively working 16 hour days through this additional work. There is some movement to change this as of late, but, this is still essentially the norm.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 1 point 1 year ago (3 children)

    Yes, but do you think they'd buy the shows from those production committees and other organizations if people weren't interested in paying subscriptions to watch them? That's like saying Bandai doesn't get money when I buy gunpla from a store like usagunplastore just because usagunplastore already bought the gunpla from Bandai months ago and Bandai isn't getting more money from that particular purchase.

    Animators being horribly underpaid is a different topic entirely.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 2 points 1 year ago* (last edited 1 year ago) (2 children)

    The people who actually made the show, animators, voice actors, and writers do not get money based on your crunchy rolls subscription, and those production committees that do get money, didn’t make the shows, they just initially financed them.

    Assuming the show is based on a manga or light novel, the original artist/writer might if they were lucky enough to negotiate shares in the production committee, but most are not in a position to do so.

    For me, what matters, is that the people who made the art get compensated fairly, that they are able to live a good life. That people are encouraged to make art by my consumption of it, and the current system doesn’t do that. It’s a horrific exploitative machine where purchase reward further exploitation of the people who actually put work and effort in to make the art.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 1 year ago (1 child)

    I don't think there's an industry on earth where it's normal for the low-level workers to be paid directly when the customer buys something. It being filtered through a bunch of business stuff is the norm everywhere I'm afraid.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 1 point 1 year ago* (last edited 1 year ago)

    Residuals are standard in the American film/TV industry. They are paid a percent of ongoing profits of previous projects they’ve worked on.

    Another fairly common practice is ESOPs where over time workers at a company receive shares in the company, and thus dividends.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 4 points 1 year ago (3 children)

    I maintain my own media library and I ensure every file has English and German subtitles. There are a variety of ways to source srt files but when all else fails a machine with enough compute can transcribe video files using open source whisper. After I generate an English srt file from the video I send it to OpenAI to create the German translation.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 1 point 1 year ago (2 children)

    Is there something similar for manga? Something that can overlay Japanese text on images, similar to what we have on smartphones but for the PC?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
    • [–] 8 points 1 year ago* (3 children)

    Although it seems likely that Crunchyroll uses an LLM for translation in some way, I wouldn't call that "confirmed" since that might be the result of an individual translator using it.

  • permalink
  • fedilink
  • source
  • hideshow 3 child comments
    •
  • [–] 4 points 1 year ago (2 children)

    The actions of an employee, when reviewed and released by a company, are the actions of that company. A company is just the sum of its employees' actions.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 3 points 1 year ago

    Also, LLM have been there for a while. So there are a few possible situations

    • LLM used is authorized or even encouraged. In this case it's the company
    • LLM use is controlled, and this falls into one of the authorized cases. Same thing really. Also their authorized use cases need review
    • LLM use is forbidden, or restricted and this is not an authorized use. In this case it falls on the company to review what's being done. It's their responsibility.

    So yeah, whatever the situation, it's on Crunchyroll.

  • permalink
  • fedilink
  • source
  • parent
    • [–] 6 points 1 year ago (1 child)

    As someone who is able to speak Japanese, I'd notice the drop in quality of translation almost instantly.

    I never turn on subs anyway when I watch my anime though.

  • permalink
  • fedilink
  • source
  • hideshow 1 child comment
    •
  • [–] 2 points 1 year ago*

    I have to since my partner doesn't speak Japanese, but half the time I end up having to correct lines for them once or twice, to make things make sense. The non-egregious stuff I don't even bother with. It's crazy how amateurish some of the mistakes are, or even what are clearly choices to omit entire sentences, for no reason.

    おい、ゆうじ君、海行こうぜ

    "Hi Yuji!"

  • permalink
  • fedilink
  • source
  • parent
    • [–] 3 points 1 year ago

    Pretty obvious if you're used any recently but confirmation is nice. Their closed captions are generally pretty terrible as well.

  • permalink
  • fedilink
  • source
    •
    [–] 3 points 1 year ago

    Both translation and subtitles have highly efficient tooling when in the hands of a professional. Translators nowadays use a mix and will build up a dynamic database as they go through a corpus that needs coherence. What's bad in this instance is not the usage of some AI, but of a badly adapted AI and ultimately of mediocre results which gives an amateurish impression.

  • permalink
  • fedilink
  • source
    •
    [–] 6 points 1 year ago (3 children)

    I feel like this is a reasonable use of chat gpt.

  • permalink
  • fedilink
  • source
  • hideshow 3 child comments
    •
  • [–] 20 points 1 year ago (2 children)

    For YouTube tutorial videos I have no issue with relying on GPT, but I think it's important to recognize that the translation of art is art. I don't feel good about the idea of something without a soul or perspective interpolating a work of art from one culture and language into another that might be wildly different from where it started.

    That all said, I think Crunchyroll and anyone else using AI art without disclosing it absolutely should be honest about it.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 1 year ago (1 child)

    I feel like what makes the most sense and is likely what's happening is that ChatGPT is being used to do the initial translation, and then a human is auditing that translation and making adjustments. So just a faster way to get the scaffolding and grunt-work out of the way.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 1 point 1 year ago* (last edited 1 year ago)

    they appear to be copying direct translations from chat GPT in to the subtitles, judging by the fact that one of the subtitles said “Chat GPT says:” and then the line in German. People who speak German also noticed that the grammar and sentence structure for many of these shows has been awful and nonsensical at times.

    If anyone is doing any sort of oversight, they don’t appear to speak German them selves and are just betting that the output will be accurate and pasting it in.

    Someone who spoke German and Japanese fluently enough to do competent oversight could probably translate faster than they could edit and rephrase the work of an LLM, which are notoriously bad at translating languages in a high context situation like dialog in a animated show. LLMs are also generally very bad with high context languages like Japanese, and even worse at translating between them and low context languages like German.

  • permalink
  • fedilink
  • source
  • parent
    • 9 comments in c/anime@ani.social view that post
    [–] 12 points 1 year ago

    They are hilariously wrong a lot of the time.

  • permalink
  • fedilink
  • source
    •
    [–] 12 points 1 year ago (2 children)

    I haven't noticed anything funky in the English subs but it's unfortunately no surprise that they'd be using AI-generated subs for languages with lower viewership.

  • permalink
  • fedilink
  • source
  • hideshow 2 child comments
    •
  • [–] 1 point 1 year ago

    It's been a bit since I stopped using CR, but I definitely noticed some AI stuff for English subs a while back when I still did. In one anime they literally went an entire episode calling a character "Brother Falcon", the entire episode was horrible to read, and it also doubled the subtitles at some point.

  • permalink
  • fedilink
  • source
  • parent
    • [–] 10 points 1 year ago

    is anyone surprised?

    after seeing subs that are supposed to say, "The basket, Amanda put it in her." comes out to, "That bitch a man put it in her!"

    happened right after Sony bought them. I swear they've also used AI voice dubs too.

  • permalink
  • fedilink
  • source
    •
    [–] 6 points 1 year ago

    I could tell recently. They’ve been seriously fucked up for Closed Captions. Only made worse by a bug recently where every episode starts with them enabled even if I have them turned off in the settings.

  • permalink
  • fedilink
  • source
    •
    [–] 5 points 1 year ago (1 child)

    I mean, AI these days pretty much always refers to Large Language Models, so if there was anything AI should and could be good for I would damn well hope it was translation.

    Like, AI gets used to make decisions, which it can't do and only emulates the sensation of. But what it does, the only meaningful thing it does, is understand language. When companies stop shoving LLMs into every crevice what will be left are its genuinely good uses like in making subtitles or helping with accessibility.

  • permalink
  • fedilink
  • source
  • hideshow 1 child comment
    •
  • [–] 2 points 1 year ago

    I think the issue is likely that they are using Speech to Text to first create a transcript, and maybe ChatGPT to edit. I dont think ChatGPT is really at fault here, its the Speech to Text not understanding names because they are from another language.

  • permalink
  • fedilink
  • source
  • parent
    • [–] 2 points 1 year ago

    Not surprised. I've seen some subs that felt off lately

  • permalink
  • fedilink
  • source
    •