Copyright holders, such as Time Warner, Disney and CBS (who helped Google test the new identification system), upload full digital copies of their content to YouTube. These copies are not distributed by YouTube at all. Rather, they are stored privately by YouTube so it will know how to recognize copyrighted content.
The copyright holders who upload their content then indicate whether they want YouTube to automatically remove uploaded copies of the same content, or whether they want YouTube to forcefully display advertisements on top of uploaded copies of the same content (revenue from which will go into the copyright holders’ pockets).
The video identification system will know when to remove a copyrighted video or display advertisements, because it will scan the frames of all uploaded videos and use a complex set of algorithms that compares them to those in YouTube’s stockpile of copyrighted material. The system won’t be perfect, of course, and poor copies of copyrighted material in particular can be expected to slip underneath the radar.
While this new technology appears to be YouTube’s biggest effort to stop the distribution of copyrighted material to date, many copyright holders will inevitably cry foul at YouTube’s insistence that they must manually hand over full copies of their content before they are given any serious protection. It’s hard to believe this effort will do much to pacify Viacom, which brought a $1 billion lawsuit against YouTube this past May.
In today’s announcement, Google outlines all of the anti-piracy measures it takes in addition to this video identification system. Users who repeatedly upload content that is eventually taken down as the result of DMCA notices are banned; hash codes of removed content are stored so that exact copies of the content cannot be uploaded again; only 10-minute long videos are allowed; copyright holders have access to an “electronic notification and takedown tool;” and users are shown “copyright tips” when they upload content.
YouTube, which has developed this new image recognition technology internally, will continue to use Audible Magic for detecting copyrighted audio in video uploads.
In large part via Webware.