Big picture still hazy in video search

Boston City Councilor John Tobin is one of the very few elected officials who publish a video blog on the Web. Every so often, he posts a short video dispatch from his district. Last month, he stood in front of El Oriental de Cuba, a great Cuban restaurant in Jamaica Plain that had been destroyed by a firebomb, talking about the support he and others were coordinating for the restaurant's owners and the residents who had lived upstairs.

But try to find Tobin's video blog using one of the major video search engines -- from Google, Lycos, Singingfish, or Blinkx.tv -- and you'll come up empty. Search for the name of the restaurant, and you still get nada. Only Yahoo's video search manages to track down the clip.

Video search is still in its infancy.

''There's tons of room for innovation," says Peter Chane, senior business product manager at Google.

Among the stumbling blocks: The search tools can't deduce much about an individual clip, such as who's talking, what they're talking about, and where and when it was shot.

Even Karen Howe, chief executive of Singingfish, a Seattle video search company founded in 1999 and now owned by AOL, says: ''This is still very much a new area. Video lags behind audio."

But a series of events, from the frivolous (Janet Jackson's piercing display at the 2004 Super Bowl) to the momentous (the recent bombings in London), have sent Internet users hunting for video clips. And some predict that video self-publishing -- much of it in the vein of what Tobin does -- is about to take off on a trajectory similar to podcasting's.

Those two dynamics, along with a desire to persuade big media companies to make their vast troves of content available, will force video search to grow up fast.

The biggest challenge today is extracting useful information from a snippet of video.

Blinkx, a San Francisco start-up, takes perhaps the most technical approach, trying to recognize the words being spoken in a video --and read any text appearing on the screen -- and turning them into searchable text.

Founder Suranga Chandratillake says Blinkx is also developing software to recognize well-known people who are speaking, both by their faces and their voices. That's helpful, Chandratillake says, ''when you want to see Bush or Rumsfeld saying something themselves, and not Bill O'Reilly commenting on it."

(Last week, Blinkx was the subject of speculation that it was on the verge of being acquired by News Corp. All Chandratillake would say was, ''We've always said we'll talk to anyone who wants to talk to us. Who knows what they'll be able to offer?")

At Yahoo, Bradley Horowitz runs the video initiative. He spent much of the 1990s trying to pull information out of videos, first as a student at MIT's Media Lab and later as the founder of Virage, which worked with government agencies to monitor foreign television stations. But now Horowitz seems to have lost faith in the automated approach to understanding video; instead, Yahoo is betting on its community of users to ''tag" the videos they watch with relevant descriptions.

''It's hard for a computer, but easy for a human, to find the funny moments of a movie, the ironic moments, the most romantic moments," Horowitz says. He cites the photo-sharing site Flickr, now part of Yahoo, as an example of successful tagging. On that site, either the publisher of a photo or a viewer can apply a tag, describing a picture of the Wellfleet Fourth of July parade, for instance, as ''patriotic" or ''parade" or ''Cape Cod."

Horowitz is also hopeful that as more video content from ''big media" arrives on the Web, it will carry some of the metadata (or descriptive elements) that were created during the production process, things like the teleprompter scripts created for a newscast.

In the future, he imagines, more video cameras will have built-in GPS chips, which will encode all video they shoot with locational information.

Google is approaching the problem differently. Google began asking users earlier this year to upload copies of the video they wanted to publish, offering them unlimited free storage and bandwidth. It also asked those sending in video to fill out a lengthy form describing each piece.

Google also created its own video player, based on open-source software. The strategy of hosting all video on its site and offering its own viewer (presumably to guarantee a high-quality viewing experience) could be a prelude to Google becoming a retailer of video content.

Chane won't divulge any plans, except to say he envisions that ''some stuff will be paid, some stuff will be free. Some will be subscription-based, and some will be ad-supported. There will be a variety of business models -- not one that applies to all video."

Chane, Howe, and Horowitz all expect video search to eventually be integrated into mainstream Web search, rather than exist as a separate fiefdom, so that a search for ''Miles Davis" might pull up websites, video, images, and audio.

''It's going to be a really tough challenge for interface designers to create something that's intuitive, and doesn't confuse people," says Howe at Singingfish, ''but I do think that's where it's going to go."

(Singingfish's software is behind the video search service at Lycos, the Waltham-based Web portal.)

But the real endgame here, for Singingfish and the others, is leveraging what they know about the world of online video to gain entrance to your television set. Chandratillake says that sites like his might be used to define a set of preferences, and have videos that match those criteria sent to a Net-connected TV. ''One of my channels might be full of news from the UK and Sri Lanka and the Bay Area -- the places I care about."

Yahoo has started something called the Digital Home initiative, about which it hasn't said much, to work with the makers of set-top boxes to bring Yahoo's services to television.

''You're not always going to want to lean forward at your PC and use a search box to decide what to watch," Horowitz says. Instead, Yahoo might help users manage an incoming stream of recommended videos from family members, or gardening videos from fellow members of the horticulture club.

Chane says users will also want Net video to be mobile, able to be viewed on cellphones, and on those increasingly popular screens in the back seats of minivans and SUVs. In other words: You ain't seen nothing yet.

Chip shot
Hiro Kataoka is on the fund-raising trail, working to put together a $15 million to $20 million round in the next few months for Boston Circuits. The Burlington start-up was founded earlier this year by Kataoka and Aaron Kurland, both alumni of NetSilicon, a Waltham company that makes embedded microprocessors.

The chip will compete with a processor called the Cell being designed by Sony, Toshiba, and IBM. Kataoka says Boston Circuits aims to ship its first products in 2007.

Scott Kirsner is a contributing editor at Fast Company. He can be reached at kirsner@pobox.com.