Today's Web search engines are driven primarily by searching for text keywords. This works fairly well when what you are searching for is words; but it fails pretty miserably when you are searching for video. Video search is almost entirely dependent on someone taking the time to manually (and accurately) add text tags to videos to make them searchable. This is all about to change, however, if VideoSurf's claims are true: The company says it has developed a way to visually search videos for it content: "VideoSurf is a computer vision search engine that has created a new way for users to easily find, discover and watch the videos from across the Web. Founded in 2006 by leading experts in computer vision and fast computation, VideoSurf has taught computers to 'see' inside videos so that videos can now be analyzed and understood. By offering a search and discovery experience powered by visual content, rather than text, VideoSurf's search engine delivers more relevant results and provides better ways for users to discover videos online. VideoSurf also allows consumers to visually navigate results to easily find the specific scenes, people or moments they most want to see, so users spend less time searching and more time being entertained. VideoSurf aims to become the ultimate destination for consumers looking to search and discover video content on the Web."
VideoSurf was one of 50 recent start-up companies to demo its products at the TechCrunch50 conference in San Francisco this week. According the Los Angeles Times
, VideoSurf appears to have made a positive impression on those who have seen it in action. Unfortunately, the list of those who have seen it working is still small. The site
is still in private beta and the L.A. Times reports that company has not even made the site available to reporters yet. The Times says that Allen Weiner, an analyst for Gartner, did get to see VideoSurf strut its stuff, and Weiner says that it is "extremely promising... If applied properly, it could yield extremely thorough and accurate information... I think it has the promise to be among the upper echelons in this space
." The Times further reports that "analysts are predicting it could become an acquisition target for one of the large Internet players
VideoSurf claims that its search engine can actually analyze every frame of a video, understand its content, and perform face recognition to identify individuals and even characters in the video. In the demo at TechCrunch50 (see the video to the left), VideoSurf CEO, Lior Delgo, demonstrated how he could find a specific clip from the TV show, Entourage
, just by searching for a specific scene. When he found what he was looking for, he was able to view the clip with playback starting just before the scene he searched for, and he wasn't force to watch the entire clip from the very beginning of the video. The L.A. Times adds: "VideoSurf says it already has already indexed an impressive collection of videos, giving users access to content across the Web including YouTube, Hulu and major television networks such as Comedy Central and ESPN. One of the key challenges facing VideoSurf will be to strike partnership deals to make available as much premium content as possible, analysts say."
If this technology works as advertised, it could represent a revolution in how video is consumed online. As the searches take place on the VideoSurf site, the company should have the opportunity to generate sizable ad revenue once the site goes public. If you are interested in checking out VideoSurf, you can request a user invitation here