Video search

By September 10, 2008 4 Comments

I’ve written quite a lot about search this year, and the general theme of these posts has been that within search there are a few promising areas for startups in which Google doesn’t deliver a great service.

One of these could be video search.

It isn’t one I’ve mentioned before because I’m a little unsure of the market potential. Yahoo! search results have ‘video’ as one of five top level search categories on their results page (alongside web, images, local and shopping) which suggests it is big, but on the other hand video is missing from Google’s equivalent list. Further, intuitively video search doesn’t feel as big an opportunity as mobile, natural language or social search.

If there is any hard data available I’d be interested to see it.

I’m writing about this today in response to a review of new video search service VideoSurf on SearchEngingLand. They wax lyrical about the power of this new service, and it does sound pretty cool. Using heavy duty computer science and a good dose of raw iron they analyse videos to pick out the most important moments and people. This is a big step up from current approaches which rely on manually created metadata or speech-to-text analysis of the audio stream. No mention of facial recognition though, but I guess that could come.

This sort of tech based differentiation could provide the basis for building a decent business in one of the smaller areas of search, which would then become very valuable for the main search players. Alternatively, if the smarts are really clever it might get bought before it gets to that stage, as Powerset was.