The last thing the world needs is another video site, right? Well, that’s not stopping Aaron Crayford, who convinced a bunch of his supercomputer engineer buddies to put together Rippol, a video discovery site that blends hardcore machine learning with social streams. Rippol (not to be confused with rrripple, the media sharing site) is coming out of stealth today into private beta. Be one of the first 1,000 people to try it by signing up here. (If you don’t get an invite, the site will be launching publicly in a month at our next Realtime Crunchup).
Rippol is still in raw form (the UI needs more polish, not all the buttons work the way they should), but it is experimenting with some powerful ideas in online video. You can sign in with Facebook Connect, or create your own account (it will also support Twitter and Gmail sign-ins soon). Once you have friends on the system, it lets you discover new videos through your friends. The homepage is an activity stream of the videos you and your friends have watched, favorited, rated, and reviewed. And a global stream in the left column shows what the broader Rippol community is watching. Links to any video can also be shared on Facebook or Tweeted out.
You can click on any video mentioned in the stream to watch it, and chat with friends via Facebook Chat, which pops up from the bottom left of the browser window. Displaying videos as a realtime stream of information is a discovery mechanism that we are beginning to see in cutting-edge video sites like Magma (our review). And I suspect it will become adopted more broadly.
However, Rippol doesn’t rely solely on your friends’ activities to help you find interesting videos. The site is also a standalone video discovery engine in its own right. You can watch about 500,000 videos right now in the Rippol player, which can be detached like a chat box and moved around in front of the screen, while you search for more videos, check out your stream or chat with friends.
Rippol ingests videos from other sites like Hulu and YouTube and then it runs them through its “Butterfly Network” (see diagram at right, click for larger image). Here is where the machine learning comes in. The Butterfly Network looks at all of your video-related activities in the site, those of your friends, and those of people in your demographic. All of the videos on the site have particularly good metadata (titles, genres, actors, descriptions), which lets Rippol take advantage of data-mining techniques to recommend videos for you. It creates a genre cloud for different categories of videos (TV Shows, Movies, Sports, News),and then generates a list of videos for each genre tag you click on.
Depending on what genre of videos you are looking for, Rippol’s realtime ranking servers shows you videos associated with ones you’ve watched or liked in the past, your friends have watched or liked, and people in your demographic have watched or liked. If you click on comedies and martial arts videos, those will rank higher for you in the future.
Since Rippol has all the realtime stream data as well, the video results are skewed towards what people are watching right now. Your own actions impact the rankings the most, followed by those of your friends, followed by those of people in your age and gender group. And if you are a 45-year-old male who tends to watch videos more like a 25-year-old male, it adjusts the demo it associates with your actual watching habits.
The system is supposed to get smarter the more you use it, and the more your friends use it. So it’s hard to evaluate how effective it will be right now with only a handful of users, but it is a promising approach. Soon users will be able to suggest videos to add to the site. (Crayford says he’s come up with a novel way to filter out inappropriate videos from the broader community by taking into account what is acceptable to different social groups).
Crayford’s last company, Vusion, specialized in high-definition video streaming over the Web. It’s patents, intellectual property, and other assets were acquired by the Clarendon Foundation in June, 2009. Craydon learned from that experience that advanced technology can only take you so far, and this time around he is trying to create a complete consumer experience rather than just an enabling technology. Rippol is completely bootstrapped so far.