MySpace announced this morning that it is offering free indexing to any content producers who want to block users from uploading their copyrighted videos. The site has been offering audio filtering and limited video filtering since late last year. The biggest question on the table is whether and when YouTube will do the same, as it promised to do when it was acquired by Google. Other questions include those around fair use of content.
MySpace has licenced the filtering technology from Audible Magic, an acoustic fingerprinting company that acquired a video filtering technology called Motional Media ID™ in November. Motional Media was invented by David W. Stebbings, a Senior VP for Technology at the RIAA from 1995 through 2000.
The company’s technology is said to recognize unique signature vectors in seconds. Sources close to another company, MotionDSP (our coverage) tell us that company will soon announce a similar product that track unique patterns of motion in video footage.
The primary technical challenges faced by these companies are scaling and variable video formats and qualities. Audible Magic’s audio filtering is said to be remarkably effective - post-lawsuit implementation at P2P network iMesh, for example is reportedly free of false positives. “We haven’t seen any false positives, and we’re seeing a recognition rate of over 99 percent,” iMesh’s founder told MSNBC last year. That was the audio program, though, not the video filter.
The political cost of trying to tame the wild west, though, is still unclear. The debate over what constitutes fair use, for example, could become largely moot if one side of the debate can lock its position down with a technology filter. Critique, satire and sampling could someday survive only with the permission of the party being satirized. That would not be good.
Ultimately though it’s unsurprising that big content owners want to control their content and existing business model instead of exploring models that are radical, new and in theory less profitable. I find it unfortunate that the enforcement of a false scarcity of content is the direction things are moving. The theory is that law and order yields investment and will ultimately mean more available content.
How popular is copyrighted content online? Conventional wisdom is that it’s what drove YouTube’s growth - but a quick look at the viral videos with the most views (via Vidmeter) and the most inbound blog links (via Viral Video Chart) shows that original content is getting plenty of attention. That’s good to see.
Marshall Kirkpatrick is the Director of Content at SplashCast and will be assisting with TechCrunch while Michael Arrington travels.
It remains to be seen how well these filtering techniques will work, and what kind of false positive to real-positive ratio they will bring, simply because nothing of this sort has been implemented on any large scale so far. If this works, perhaps organizations like the RIAA will lay off tech companies who, if anything, give their content more popularity, and go back to the stone age of selling CDs with 1 good track and a bunch of filler items.
I’m really interested to know how successful this kind of technology can be, there’s surely got to be false positives with this kind of thing?
Thanks for your comments, I added a few lines about false positives.
This simply won’t work - bottom line.
Whether or not the video classification is super reliable, this announcement definately will put pressure on YouTube to do likewise, raising their operating costs (e.g. dedicated servers running the detection code) and possibly somewhat discouraging their customer base (longer submit times).
While video is just a portion on MySpace’s offering, it’s all that YouTube does. So any changes to their processes will have a huge impact on their product/experience.
I’m skill skeptical on the video classification, it’s a very hard problem to solve in a robust manner (i.e. identification works across codec compression and other means of modifying the original). So I’ll be following this one closely.
Very interesting post!
Good to see that original content has a serious share.
Youtube has also declared that they’re going to pay users for their content, something that Metacafe and Revver already do.
Meanwhile they’re distributing content to Verizon customers for a monthly fee, and their producers are still left unpaid.
I would be interested in seeing how this system factors in “fair use” applications of copywritten video. Just because the system could recognize a ten-second music or video clip in my video, does not mean that my video violates any copyright and should be removed from the site.
Oh, right, I forgot that the RIAA and MPAA think that “fair use” is a legal concept developed by the Tooth Fairy and the Easter Bunny while interning for Santa.
How long before some 12 year old finds a way to hack this new technology so that folks can work around it.
I give it about a week.
This is where all the online video hosts are moving towards. They are investing heavily in research and development into better and better programs for detecting these copyrighted materials…
The revenue potential alone from all this usage data puts Audible Magic in a very sweet position.
I’ve had a few videos blocked on YouTube and was told it was because I was using copyrighted music. Actually, I was using music that was written specifically for my shows, so whatever system they have is hit and miss. It’ll be interesting to see how the MySpace thing shakes out, because someone will eventually get uber pissed when their particular content gets dumped.
The scary thing is that MySpace is way too big to handle the individual needs of content creators in situations like this, that is why sites like Revver and Blip are desirable for those who are spending more than 5 minutes putting together a video.
There are a number of video editing mashup web sites and others launching soon that allow consumers to take media from who-knows-where and create hybrid edits incorporating sound, images and video from these multiple sources… I’m curious how this filtering will handle that. I’ve seen an example where the theme song for 90210 was used to support video of a group of friends interspliced with cuts from movies showing scenes from LA. In total, they had about 6 different films, plus their own video and that theme song edited into a 30 second piece. Is that fair use? Would this filter block that?
Hey Marshall,
Does that mean you’re back at TechCrunch? Or maybe you never really left?
- Steve
I think it’s a smart idea to create a tool and let the companies use it themselves. This is a really huge issue in the industry right now, but I think the user follows by example when it comes to social networking sites - they sort of tend to do what you lead them to do (so if it’s posting sexy pictures up, that’s what you’ll see, or with our site, taking a picture of your outfit in the mirror…) - it’s not to say it’s an easy fix but I think the consumers aren’t nearly as concerned with whether or not they can use stuff as the companies are and that’s where the issue is.
Yay Marshall! Nice to see you posting again!
making sure i can post
did i get this right? so basically a content producer or one with the rights to the content has to go thru the steps of indexing the content with myspace so myspace can tell if the content is being played on the network w this new technology. doesnt seem like a bad idea for myspace to have all the biggies working on this with them. also seems like a pain for someone to have to do that and adds in a ton of extra work for them as well as the aggravation of losing potential revenue bc of people uploading copyrighted material to the sites.
i’m def interested in hearing more about this type of copyright detection software and other firms that are trying to do it. seems to me like a big oppty for the person that can figure it out.
also the issue of mash-ups and what constitutes one vs a stolen piece of work will be very interesting to see how it plays out. Im a big fan of mashups so would be sad to see enforcement against them.
all the video sites will inevitably be forced to have a revenue share model, or to implement this copyright technology - in order to please everyone across the spectrum
for the most part, it works - you guys are pointing out the exception to the rule and making it appear(potential ways around it) to be a problem of huge magnitude
This is big trouble for YouTube - the content filtering doesn’t even have to work very well, it’s enough that it exists in some form and YouTube aren’t implementing it. If any copyright violations went to legal actions, YouTube would seem to not be doing enough - as Myspace have ’shown’ that copyrighted content can be blocked through software.
I agree with Will - this is a wakeup call to Google. Big media companies aren’t going to accept ‘we’re working on it. we swear!’ anymore - recent moves from Viacom and NBC certainly reflect that.
This is just icing on the cake.
Good questions about false positives (and Negatives).
My Motional Media Identification System is powerful for the reason that the false positive rate can be dynamically set to be any value that the user (MySpace in this case), feels comfortable with. For example with a million programs in the library it can be set to have an error probability threshold at -say- one trillion to one, less or more at choice. The only condition is the duration of the clip being analyzed. Falsely measured vectors are ignored should the clip be so distorted that erroneous measurements were made. In such a case the process continues to compensate for the error. So far on really terribly distorted images on UGC sites there have been zero false positives and zero false negatives. Just because a clip is identified does not necessarily mean it will be blocked, that is the perogative of the owner of that content, in fact some creators like the publicity gained and consumer’s enjoyment, by letting it go out.
There is a company - Vidyatel Ltd - that is working on video fingerprints for the last 10 years. They implemented 10 years ago an editing suite and Real time TV broadcst monitoring service.
The system is works well. As I know it works in real time, with no problems of false alarms.
Whether it’s Vidyatel or Motional Media ID, such applications must be deployed with care, to promote content publishing.
Vidyatel’s solution is very compact and can identify the content with a very short fingerprint.
I do agree completely with Ishai’s comment that such systems must be deployed with care, not just because of the technical issues but more importantly because of the consumer and creator factors that must enhance the furtherance of education, enjoyment, promotion and quality of life of all the people involved.
Technically:
Motional Media ID is compact. In fact only about 8 to 16 bytes are needed to identify the program from a clip, if one is wanting a trillion to one probability of making a correct ID. The database for 2 million hours of video requires only about 1.2 GigaBytes of fingerprints. Therefore searches are fast, and the software analysis and logging system can run at multiple tens of times real-time, in fact as fast as the video can be supplied, whether from a hard drive, Internet ultra high bandwidth, or of course even much faster from RAM. (Assuming a today’s PC processor). There are other considerations and aspects of the power of the system such as its inherent robustness that are beyond the scope of the discussions here.
Your invitation to play LOST:
http://www.lost.eu/20ac1