TV Guides for the Web

3 of 5

VideoSurf

VideoSurf organizes content based on visual recognition of faces and images in videos, rather than using text-based data such as meta tags. The technology returns a string of snapshots showing frames within a video, much like fast-forwarding through a DVD. Each query generates a visual list of people or characters; from there, a user navigates to videos in which the subject of the query appears. The technology filters out subpar or incomplete videos. VideoSurf trolls content from more than 10,000 sources and has indexed more than 480,000 TV episodes and 25,000 movies. VideoSurf gets revenue from sponsored videos.

MONTHLY UNIQUE VISITORS: 1.5 million