Hacker News Reader

Comment by smnscu | Hacker News Reader

It'd be interesting to build a tool to detect all the copyrighted material on YouTube that bypasses the automated checks. Some of the tricks used:

* picture-in-picture: https://www.youtube.com/watch?v=Rp1aOWUSRZg

* mirrored image

* higher pitch

* other?

View on HN

Interesting in what way? Everyone who has spent more than five minutes on youtube knows their bread and butter is copyright infringing videos. It's the unspoken dirty secret of online video.

Just search for 'full album' or 'full movie'.

Beyond the outright movie/album piracy, there's another phenomenon that happens constantly at Youtube:

1. Someone uploads funny/interesting content.

2. Other people (bots maybe) notice that it's popular and download and re-upload it, titling and promoting it so that it ranks above the original in search results. Often the re-uploaders file takedown notices against the original video or other re-uploaders, so they're removed.

3. Repeat step 2 hundreds of times, until any search for "funny/interesting content" returns only a series of almost-identical 240p 15fps cropped, mirrored, or otherwise unwatchable versions of the original video.

I don't know if they condone this because it results in more overall views, or if they're just terrible at fighting it, but it's largely made Youtube useless for its original purpose as a many-to-many video sharing platform.

Hey CommieBobDole, that's exactly the problem we are solving. Our technology is robust enough to detect all those tweaks.

It's hardly unspoken, Viacom sued them for a billion dollars way back before Google bought them. They only lost the case because they couldn't find evidence that YouTube execs mentioned any specific video when discussing the money they made from piracy.

They fact that they included videos they had specifically authorised or uploaded themselves in the list of supposedly infringing videos couldn't have helped.

One correction: Google bought YouTube in 2006. Viacom sued YouTube in 2007.

I guess that makes more sense, since YouTube wouldn't have had $1B to sue for. But the actions brought up in the trial were pre-acquisition.

Other tricks I've seen:

* skip a 1-3 seconds every so often

* chop it into blocks and re-arrange them, so you watch out of order but still get the gist (1,2,3,4,5 becomes 1,3,2,5,4)

Hi robbiemitchell, we can also recognize video with that technique :) You should try the platform

Yeah, those are fun ones. We [0] are dealing with most of those.

Here are some fun ones based on Shia LeBouf's green screen performance https://www.youtube.com/watch?v=ZXsQAXx_ao0

- https://vk.com/video-67185996_171327939

- https://www.youtube.com/watch?v=w_TuTzzSUb0&t=0s

- https://www.youtube.com/watch?v=i4ktEzJvaGM&t=186s

- https://www.youtube.com/watch?v=5BLMW3vQV20&t=124s

etc. There is much more we can do, but there are some we can't. For instance phone recorded 3D movie recorded in a cinema from a weird angle.

[0] https://pex.com

I'd be more interesting to build a tool to detect all the videos in Youtube that fall under fair use but are still being DMCA'd.

We're looking into this for couple of years now. We're doing some steady progress, but far from a final result.

The problem is incredibly complex, because Fair Use doesn't have any quantifiable metrics that you can translate into a computer algorithm. The gist of Fair Use is that you have to prove the intent of the uploader trying to rip you off.

As you can imagine, this is a very hard thing to prove and the amount of videos being uploaded every day makes this is even harder to crack.

If you're interested in this problematic, reach out and I will be happy to share more.

Could be a cool open source project. Would almost definitely be a worse business.

Thanks for starting a good discussion smnscu. That's exactly what we are here for: we are solving that problem!

How will you deal with fair-use?

Since there is no algorithmic way of defining fair use, it can only really be dealt with by lawyers in court.

Hi abdias, thanks for your question! At this stage we only inform our users where the uploaded video is online. Do you think we should deal with fair-use?

You don't have to, fair-use is a defence only applicable in court, the easiest way is to sue everyone and let their lawyers assert fair-use. AFAIK it's not unlawful, nor will you incur any fines, if you sue someone for using your copyrighted material if it ends up being declared fair-use. Additionally, I think youtube is even more lenient than that towards people asserting unlawful use of their copyrighted material. In short, fair-use does not exist outside of the courtroom and no amount of "I own no copyrights" or "No copyright intended" tags, or citing the copyright code on your videos can summon it.

YouTube already does an absolutely horrible job with handling take-downs and fair use. I have zero interest in seeing any systems or algorithms built that would aid the sloppy, lazy, and greedy organizations shot gunning take downs, even on their own material and channels (always funny).

This amuses me greatly, got any tangible examples?

I posted an unlisted video of my daughter at a noisy indoor theme park. It got automatically removed because the venue had a song playing in the background.

jwz posted a thing about his long saga fighting a takedown of horror movie reviews.

>even on their own material and channels (always funny).

I was referring specifically to this? Am I right to infer that DMCAs are so scatter gun that publishers have managed to DMCA themselves?

>Am I right to infer that DMCAs are so scatter gun that publishers have managed to DMCA themselves?

Yes this has happened multiple times. Notable one that's somewhat recent that comes to mind: https://torrentfreak.com/warner-bros-flags-website-piracy-po...

Your line of thinking is why we need anti-SLAPP style laws protecting Fair Use.

Strategic lawsuits against Fair Use are a form of stifling protected speech.

> sue everyone and let their lawyers assert fair-use

YouTube would turn into a tumbleweed landscape overnight!

I've often thought YT should just add "No copyright intended" et al as a ContentID trigger though.

"No copyright intended" always makes me laugh. What does that even mean?

It means the average person has no idea how copyright law works. It means lots of average people think there is nothing wrong with non-commerical usage of copyrighted works so long as you don't try to pass something off as your own. It means the legal definition of copyright infringement is out of step of what most people think of as right and wrong.

Cell phone video of a television.

Hey emodendroket, that's one of the use cases we are better at: recognizing videos from a scan of the TV.

Reply on news.ycombinator.com