YouTube is shutting down crowdsourced captions
arstechnica.com
arstechnica.com
This is kinda like a restaurant removing their handicapped ramp because it's "not used often". Of course, that'd be illegal.
They should think twice about bringing this one back. There may be some extra burden on their end too with violation/sabotage contributions showing up to contend with as well.
There is no way automated captions can be anywhere near OK for that.
He currently has a community of volunteers supplying captions and has as a result an audience of Russians and other eastern europeans, who partly used to drive or work on these tanks, watching his channel. This is now made impossible thanks to google.
They aren't going to change their behavior - this is how Google rolls and it works for Google. They only want popular, cool stuff - and catering to the disabled and less fortunate isn't popular or cool.
I doubt much can be done about this... Try to avoid their services, when possible? Don't give them money? Encourage others to avoid Google?
It's frustrating having no recourse.
While the former is certainly true, the latter is definitely not. It's literally in the mission statement.
In this case, kind of, Google does work pretty hard on accessibility features. No one else has workable live transcriptions, and android accessibility suite is seeing constant improvements.
Maybe they think their ML transcription has gotten to a point where the process can be automated and maintaining crowdsource captions is not worth it / has liability. Still I disagree with this move. Genuinely curious what the cost/benefit analysis is. Most countries demographics are skewing older, I think it's worth maintaining these systems just in case.
When they turn around and cut off a surely much less cost-heavy accessibility feature seemingly at random, to many users complaint, it is, in fact, surprising.
As far as recourse, it'd be a bit of a stretch to use the ADA here, but I wouldn't be totally surprised if someone tried a lawsuit in this instance: Google is cutting off handicapped users for almost no real cost to themselves.
Yes, that show includes a variety of UK English, Irish, Scottish, and Welsh accents, but... damn, it's bad.
It's just not a smart move.
and
it's too expensive to roll out your own personal solution and share it with other people who think like you(think global scale not acquaintance scale) without using big tech
It's not quite like removing a ramp. It's more like a city had a program of "we'll help pave and maintain the sidewalk in front of your business to be more accessible, and to add ramps to connect to your door, by request"; but then no businesses took them up on that offer by requesting ramp installation—so they stopped offering the program.
Youtube is essentially a Monopsony for monetized Internet video. So this will make many previously accessible channels inaccessible to the deaf and hard of hearing.
There is a pretty fine line between cynicism (i.e. it's just marketing) and idealism (ethical companies will generate a higher level of support), but I wonder how far this goes before a lawsuit shuts it down.
It's not unreasonable to think pushing for an accessible OS makes the rest of it better as well.
One nitpick, I would disagree that accessibility features will always be low usage. Many accessibility features have become ubiquitous and essential features of many products. And many people preferred such captions who did not identify as needing accessibility features.
https://support.google.com/youtube/answer/6373554?hl=en
Creators can also add their own captions in multiple languages, but this relies on them working together with the community, which may not always be realistic.
I can imagine how invaluable it is for a deaf person, they must get used to assume every transcribed word has only a 60% probability of being right, and quickly infer an array of possible meanings through context. But the experience is not comparable to human-transcribed subtitles.
completely devoid of punctuation
and proper capitalization so that all the words
run together in an endless nonsense sentence also they're
useless because each word pops up on the
screen one a time very slowly in time with the audio
with arbitrary line breaks sounds useful but
it makes it impossible to read the captions and actually
watch the video because you're forced to keep
your eyes on the weirdly scrolling captions you
can't just read one line at a time like proper
captions and it never gets the hard to
hear words right you know the ones that
would benefit most from proper captions
---
Also, the machine-generated captions are completely devoid of punctuation and proper capitalization, so that all the words run together in an endless, nonsense sentence.
Also, they're useless because each word pops up on the screen one a time, very slowly (in time with the audio), with arbitrary line breaks. Sounds useful, but it makes it impossible to read the captions and actually watch the video because you're forced to keep your eyes on the weirdly scrolling captions--you can't just read one line at a time like proper captions.
And it never gets the hard-to-hear words right. You know, the ones that would benefit most from proper captions.
Creating real subtitles is as much a combination of art and skill and judgement and balance as editing the video itself.
A good subtitle allows you to quickly read ahead, and then move your eyes back to the video, back to the people speaking, so that you can watch them act out the line.
A good subtitle doesn't transcribe everything said word for word, but reduces spoken language to written language, and focuses on the gist of what's said.
A good subtitle for hearing-impaired viewers describe important sounds that happen off-screen, such as doors opening or closing, but skips non-important sounds, and sounds of things happening on-screen.
Good subtitles and CCs both have to be structured carefully so that they do not negatively impact the delivery of lines in the content. The biggest offender of this is comedy series as an improperly structured and timed CC or subtitle will ruin many jokes.
Another aspect of this is how they handle making it easy to digest multiple lines of dialogue. Good CCs and subtitles will either structure the dialogue such that a certain voice is always on top or bottom or alternatively will match colours to voices.
Similarly, in content that is visually busy or rapidly changing, good subtitles and CCs will stay out of the "busy" section of the screen and avoid using a colour or outline that would interfere with the readability based on the background of the scene.
They are different but with regards to what makes them good, they are basically the same.
And to add on to this, on just about everything but live feeds, they are produced and added to video more or less the same way.
Yes. The rules for how long to show a subtitle, and when to pop them in and out are the same, regardless of the actual text shown. Everyone can decide what characters-per-second limit works for them, but whatever you choose, it fundamentally alters everything else and restricts what text can actually go in the subtitle. Translated or not translated, sound cues or no sound cues.
And in any language, the machine-generated captions don't have punctuation. You also miss out on non-speech cues that a manual captioner will often add, like a note that there's music playing in the background, or that a loud crash happened off-screen, or simply that a person laughed.
> Machine captions can be scaled to caption the entirety of YouTube, something crowdsourcing could never accomplish.
They certainly don't scale the same way, but one of the entire purposes of crowdsourcing is that it scales better than trying to employ what will inevitably be a smaller number of people. I still think there's value there, at least until machine-generated captions are actually on par with a human. We're not there yet.
iOS seems to have a lot of accessibility features, and without impairments I still find them useful.
Captions are great, who wants to listen to crappy audio all the time when one can read 5x faster?
This is statistical illiteracy that even smart people continuously use. It is absolutely wrong!
Starting from August 2019, only uploaders can approve submissions, when previously other viewers or YouTube moderation could also publish. A lot of people seem to have forgotten this, and YouTube never really updated the UI/Help pages to reflect it.
I have to wonder if whichever analyst came up with those stats was aware of this but turning a blind eye, or simply never noticed.
Most videos I watch are on niche and technical topics and the auto generated captions do a remarkably poor job, so I will miss this feature very much. RIP.
That's because the spam is a problem the content creator has to deal with. According to the article, captions need to be approved by the creator before the public sees them, so you'd never know.
Removal of this feature is going to impact our international students very badly. Access to technical education was already hard enough - now it'll require English fluency too.
YouTube isn't removing existing subtitles or the ability to have subtitles. You'll just need to use a different platform for people to contribute them, no? And then the creator will still be able to add them to YouTube.
Ideally there won't be any negative impact on international students at all.
So I really do think this will be a net benefit in the long run.
1. Can out-of-the-box VLC be an adequate replacement? It can play YouTube videos and it can also load & display .srt caption files (least work for me, but will hit channel's incomes)
2. Can browser extensions easily query the playback time on a YouTube video with sub-second accuracy?
3. Can a browser extension easily overlay text over a YouTube video? I'm thinking anti-clickjacking security might get in the way, but I'm not familiar with the current extensions security models.
https://chrome.google.com/webstore/detail/subtitles-for-yout...
Your system can close caption clear sounds on a good quality recording over no background noise? awesome. That's exactly what I don't need it for.
It also lets them avoid a lot of difficult decisions wrt policing speech which is something youtube has established a long history of being awful at.
Anyway, I know you're not apologizing for them. Google already has 125k+ employees. They could hire 10k+ more, it wouldn't really matter.
This is such a bad look, what were they thinking?
All adtech and most social media is both abusive and socially destructive, but it's profitable enough for the Right people that there's no serious prospect of it being shut down.
The road network sees a lot of abuse (e.g. transportation of illegal property, impaired driving, simply allowing criminals to move freely) but it's too socially useful to shut down.
Fear of abuse is just a cover for withdrawing, or failing to provide, services for other, unstated, reasons.
(I am hearing, but like clean english captions because one can easily play at 2x or beyond.)
(Is it really so hard to believe that the usual neural network progress curve has happened for speech transcription and that the future is already here, just unevenly distributed across videos?)
Not saying I agree with Google's decision. Just saying it makes sense that you wouldn't see spam and abuse.
If the article is accurate -- that crowdsources translations always had to be approved by the creator anyways -- then I don't see why third-party sites for this won't pop up, which any channel can direct its users to.
> YouTube says it's killing crowd-source subtitles due to spam and low usage.
I mean, while the feature was cool, it honestly makes sense to me that it was too niche for YouTube to focus resources on maintaining. (And a lot of people drastically underestimate the constant attention spam prevention requires -- it's horrible, really.)
Seriously -- this seems like an awesome opportunity for someone (anyone) to create a superior tool for this, no? A separate community with quality control tools, reputations, discussion boards... And then the creator just easily imports a finished crowdsourced sets of captions/subtitles to their video.
Honestly this feels like an ideal "Show HN" project. Not only that, it could grow to become a community hub for crowdsources captioning/subtitling for videos across all platforms, and even bootstrap itself into becoming a major video hosting platform in its own right.
Maybe I'm just a glass-half-full guy here, but I see this as a real opportunity. And I have to assume a third party site for this will ultimately do a better job than Google's tool enabled anyways.
I'm sympathetic towards YouTube on this particular issue. I clearly remember that there was an early time when a lot of slurs were filtered from YouTube's auto-generator output to prevent such incidents, and then there were some furious complaints about how YouTube censors words in the caption. For example, see [0], and note how the comments are overwhelmingly negative, "Is YouTube becoming a kindergarten now?", "Haven't you got the latest dictionary of Newspeak?", "The age of snowflakes", yet no one realized the serious technical problem of incorrectly identifying words as slur.
And now many word filters on auto-generator output have been lifted as the system became more reliable, then we get some furious complaints about how YouTube incorrectly identifies words as slurs.
I think there has been a similar accusation that claims Google removes negative keywords for certain politicians from its search recommendations. But a quick test could confirm that it was true even for convicted criminals. Same reason - incorrectly recommending negative keywords about living persons would be a serious incident.
[0] Youtube are Censoring the word BULLSHIT https://www.youtube.com/watch?v=bHvkEPL7mVA
BTW, I don't know whether the word filter was used on all subtitles or only automatic ones, so I'm limiting my scope of discussion strictly on automatically-generated subtitles.
For example, imagine there is a video where someone says a slur and the system transcribes that slur without censoring. If you detected it correctly (true positive), then everything is fine. However, if there was no slur and you detected it as a slur (false positive), it is a big problem. Imagine watching some video talking about perfume, the narrator says something like "the smell lingers for a while", and the transcription code thinks it got the n-word instead of "lingers". It would be disastrous.
Now imagine there is a censoring system in place for slurs. What it does, it simply gets rid of false positives at the root by completely removing them from the list of possible words to autogenerate, thus guaranteeing that situations like the one I brought up above could never happen. So now, if it is a true positive that nets a slur word, it just gets rid of that word. PR disaster avoided.
"Censorship", in this case, is just a very unfortunate side-effect of solving that problem I described above. As long as they don't censor those words in manually created captions, I am totally ok with it, because then the burden of a false positive is on the person who wrote the captions. I.e., if google doesn't ban "true positive slur" scenarios in manually generated captions, imo it is a reasonable and understandable compromise.
Autocorrects solve this by omitting the word from their dictionary altogether, so if a false positive happens it's from duck to luck, not duck to something censored.
Voice: "Because lucky you, that's why"
Caption 2: "Because f*ck you, that's why"
Caption 3: "Because [censored] you, that's why"
Caption 4: "Because [unintelligible] you, that's why"
You tell me, which one of these failure modes is better?
That would solve both problems at once, because there is a really strong chance that the first closest guess that isn't a slur is the actual correct word that was said there. And in case it picks an incorrect word, it would have picked it either way and it would have been just as wrong.
So if the voice said "Because lucky you, that's why", in case youtube autotranscript thinks it was "fuck" and not "lucky", it would either default to "lucky" (which is good) or another closest non-slur word like "mucky" or "rocky".
And, in my opinion, this is way better than putting [censored] or any other workaround out of those you mentioned, because all of them have zero chance to end up as the correct word (in case the slur was detected incorrectly), while with the approach I described, it is very possible.
Caption 5: "Because you, that's why"
Not the best failure mode, but it does completely prevent the damage, <del>and your argument about how censorship doesn't prevent damage doesn't apply</del>.
https://www.theguardian.com/technology/2016/dec/05/google-al...
I still believe Alphabet is a company that truly wants to make the world a better place -- I hope they're living by their principles, and not letting the profits (of which they have plenty) get in way.
> because many of you rely on community captions, YouTube will be covering the cost of a 6 month subscription of Amara.org for all creators who have used the Community Contribution feature for at least 3 videos in the last 60 days.
What in the world is their motivation here?? This is the kind of thing a company wouldn’t do if they were worried, at all, about competition.
Wat
I mean, Google kind of are a video captioning company. Google Video was created with the goal of letting you search videos by indexing the automatic textual transcriptions of their audio content. YouTube still performs that role for Google Search; that's why Google's automatic transcription is as good as it is already.
1. You see something you want captioned
2. You go to that video's Amara.org page
3. You caption the video there
4. The video creator exports it from there to Youtube.
Also, it appears creators can still write captions. Honestly, this doesn't seem like too big a deal.
Because, if a website can embed a YouTube video of a Windows desktop; and either use Javascript to make it fullscreen, or convince you to make the video fullscreen and then "animate" the video closing itself to give you the illusion that you're no longer in the video; and then use annotations to turn the Windows desktop video you're now staring at into an image-map, such that the user will click on the video, thinking they're closing a window on this "desktop"... then you can get all sorts of bad things to happen.
Remember: many websites (e.g. blog hosts, forums, commenting systems) that don't support Javascript per se, still support automatic YouTube video embedding by posting a YouTube video URL.
How could the attackers video simulation of a windows desktop fool the user into thinking they are still on their own computer?
To such users, an OS desktop is an unpredictable thing. Their (live) wallpaper and (dynamic) lock-screen background change out from under them at random. Their theme changes between Light and Dark (and they haven’t yet noticed the concordance with the time of day.) The Google Doodle changes without a seeming pattern (and they’ve never clicked on it to find out why.) Every once in a while, an app updates and looks entirely different after a restart. New application shortcuts appear seemingly at random on their desktop. Whenever their nephew “fixes” their computer, most of those disappear, and many other features get changed as well. Was it something the nephew did? Or were those just temporary, and disappear if you restart the computer? Or were they viruses?
Why would such a user even be surprised if, one day, the computer decided to change everything about their desktop out from under them while they were in the middle of watching a video? Par for the course. Computers are unpredictable.
—————
But also, it’s pretty easy to make the “desktop” part of the desktop irrelevant. The goal here is to re-invoke the concept that the user is looking at their OS chrome (i.e. “windowed mode”, rather than “fullscreen mode”); not that they’re looking at their desktop specifically.
So, in the video, just display a maximized Chrome window showing YouTube navigated to the innocuous “bait” video that was previously playing up to that point, as if the video had been first navigated to and then de-fullscreened. And then, show the pop-up window on top of that maximized Chrome window.
The only thing the video’s author needs to get right, at that point, are 1. Chrome’s window decorations, and 2. the taskbar.
(It doesn’t matter if the user isn’t using Chrome; even I, as a power-user, wouldn’t find it strange to find that something I had done had managed to install Chrome and force it to open a URL. Happens on Windows with things accidentally opening in Edge all the time. And lots of programs have dark-pattern installers that try to install Chrome for you; especially those of other Google products, e.g. Google Earth, Google Drive, etc.)
It's even worse than that.
I use Firefox predominantly, but I do have Chrome installed for the occasional use (webdev, some browser-discriminating webapps, more stable Netflix). I'm a power user too, and yet I already had a few situations in the recent years, where I'd switch to a browser to find some information quickly, only to realize after few minutes that I'm using the wrong browser.
And for those tech-scam-vulnerable people, a browser is just a big rectangle showing the Internet, with a bit of visual noise surrounding the tab bar and the box in which you type "google". I know some people who probably wouldn't notice if I switched their Chrome for Firefox, or perhaps even for Edge.
I've got no hard evidence of this, but I feel like the removal of that feature got rid of a lot of the annoying spam, and the people who were using it correctly just started paying more attention to their videos before they uploaded them.
Obviously some will still get missed, but I think the overall accuracy of the actual videos has increased substantially.
Google is doing no different here. They are not getting rid of captions. They are not getting rid of user captions. They are only getting rid of their onsite caption editor. You can still create captions a thousand other ways. You can still give them to the author of the video and they can upload them into youtube in supported formats and apply them to the video.
Want to community edit a captions? Google docs or Google sheets would be two fine places to that. So might github. There's no need for a UI on Youtube for this.
> YouTube says that because "many of you rely on community captions, YouTube will be covering the cost of a 6 month subscription of Amara.org for all creators who have used the Community Contribution feature for at least 3 videos in the last 60 days."
That arguably points out 2 things. (1) they aren't abandoning disabled people and (2) they're probably not lying about the feature not being used since if it was being used this offer would cost $$$$$$$$$
It sounds like nothing really changed. You can still contribute subtitles to a channel, one of the few that actually uses them, and the channel creator can still attach them to videos. All that changed is the interface to edit is no longer in youtube. Choose any of 1000s other ways, send the creator your subtitles, creator can attach them to video.
Given the effort of making captions is actually the work of writing the captions themselves the effort went from several hours of work to the same several hours of work +1 minute. Literally nothing of importance changed. And that +1 minute assumes Youtube's UX was best in class because of it wasn't then moving to a better UX could save more time.
The level of animosity toward anything Google does is absurd. If there was a captioning service making money and Google added captions they'd get shade for ruining a market. Now they are removing an arguably rarely used UI but still leaving the feature (captions) completely available. There's literally no harm. The people actually motivated to write captions will barely notice the change. They'll still write them. All they have to do is send them to the channel creator. If the channel creator doesn't want them nothing has changed. They had to approve them before, they have to approve them now except click 1-2 buttons to upload them. Sheesh
The process before was straightforward: a community member would author captions and submit them to the producer, the producer would choose whether to approve them right there on YouTube's site; the new workflow involves email, forms, files, and community members who are motivated to go out of their way to figure out how to create and communicate YouTube-compatible captions that will now be less likely to be published even if they are high quality.
And that means youtube is entrenching creators into their homemade economy even deeper. Already it's a fantastic phenomenon that people have created careers out of youtube.
But they are 100% dependent on Youtube.
Creating a new expenditure for them to have subtitles will only make it more exclusive and make them even more dependent on their youtube salary.
Small independent creators will have no chance to compete.
I have exceptional hearing and have no learning disabilities and I use captions in videos all the time.
I actually think this a good thing. Maybe the community will actually we that dependence on proprietary systems isn't a good thing.
They already exist, and the subscriber mechanism is already better than YouTube (view offline, choose when/what to download), just needs some work for comments, ratings, annotations etc.
It's somewhat amazing Google and YouTube don't do it.
So I'm not sure what the real reason is here?
I could see spammers flooding shitty videos no one watches with fake captions, or popular videos with obscure languages and fake captions and it destroying their search. But currently it doesn't seem to matter anyway.
On the bright side, dozens of browser extensions and scripts will have to be created to solve yet another needless problem created by Google.
Does it? Why won't anyone answer?