With opportunities as endless as the millions of videos online, there is no shortage of ideas in the wide-open world of online video advertising.

Some companies use speech recognition software or automated visual analysis to match unclassified videos with relevant ads. Others “wrap” ads into the viral videos themselves, while other upstarts create personalized video ads on the fly.

The developers behind these so-called contextual targeting technologies hope to categorize millions of user-generated videos—which far outnumber professional movie or television clips available online—and match them to Internet ads. The aim is to deliver highly-targeted video ads that are relevant to the content people are watching at that moment, and the startups that figure out how to do this expect to be handsomely rewarded.

But a number of digital marketing companies contend that big brand advertisers are far more interested in people’s demographic profiles—typically determined by Internet cookies that track their web site visits—rather than contextual data drawn directly from the specific videos that people are watching at any particular moment. That’s why a beer producer trying to reach older men is much more likely to sponsor football programming, rather than an Internet video of a bunch of college students getting drunk.

“Branding is 90 percent demographic, 10 percent contextual. When doing branding it’s about who is watching, not what they’re watching,” said Iggy Fanlo, CEO of San Francisco-based online advertising startup AdBrite.

But that hasn’t stopped dozens of startups from trying to bridge the gap between large advertisers and user-generated content. These companies believe that the unruly world of user-generated video is where the biggest future opportunity is and are trying to plant that flag.

Video ad spending is expected to make up only about one percent of the amount U.S. marketers spend on television commercials in 2007, according to eMarketer. However, online video ad spending is expected to rise from an estimated $775 million this year to $4.3 billion in 2011, representing almost 10 percent of television ad spending, according to eMarketer.

The cast of companies includes ScanScout, a Cambridge, Massachusetts-based startup backed by General Catalyst Partners that uses a combination of audio analysis of the speech and music as well as visual analysis of patterns in the video to precisely match it with a relevant ad.

Rivals such as adap.TV and Digitalsmiths—backed by Aurora Funds and Chrysalis Ventures–also do such matching, known as contextual targeting. There are various ad formats: adap.TV—backed by Redpoint Ventures and Gemini Israel—has “inline” text ads on the bottom of the video with hyperlinks to advertisers’ web sites, while others have text ads on the bottom of the video that, when clicked, pause the video and expand into video ads.

While most companies insert ads into videos when a user goes to a web page, Kiptronic, a former podcasting company backed by Blueprint Ventures and Prism VentureWorks, focuses on downloadable videos. The company inserts ads on the server side so that users take the video and ad on their iPods or wherever they go.

Sequoia Capital-backed AdBrite “wraps” ads into videos so that wherever that video spreads on the Internet it has the ad inside it. Cambridge, Massachussetts-based EveryZing, an audio and video search engine funded by Fairhaven, Accel, and General Catalyst, focuses on transcribing speech from publishers’ videos to better match it with ads.

Others take a radically different approach. Australian startup Qmecom offers personalized video ads targeted to specific geographies, genders, and ages. New York-based Visible World has a similar product.

While many companies focus on particular parts of the process—such as serving, inserting, matching, targeting, buying, and selling of video ads—Westminster, Colorado-based SpotXchange, a division of digital marketing firm Booyah Networks, aims to be an end-to-end solution.

SpotXchange, launched in November 2006, allows advertisers and publishers to buy and sell video ads in an auction format. In the vein of Yahoo/Right Media’s display ad exchange—but for video—advertisers on SpotXchange can purchase ad space from large, medium, and small publishers and target particular geographies and demographics.

These startups hope their interest in user-generated content will give them an advantage over larger digital marketing firms such as DoubleClick, eyeblaster, and aQuantive, which are geared towards matching ads with professionally-produced content.

“The market is dominated by high-end brand advertisers and high-end brand content with a high sell-through, sold-out situation with leading major publishers,” said Ari Paparo, DoubleClick vice president, rich media. “As soon as you go to the mid or long tail there’s massive undersold inventory where buyers aren’t interested in that business yet.”

Most of those advertisers have so far shied away from user-generated content because it is entirely unpredictable. Homemade videos have proven to be a hit among Internet users, but advertisers rightfully point out they don’t want their ads or brands associated with violent or offensive videos. In the most recent example of a user-generated fiasco, several major British advertisers pulled their display ads from Facebook last Friday after their ads were placed with a Facebook user-generated group for white supremacists.

But there are signs of progress. Startups such as Redwood City-based YuMe Networks are going after the so-called “middle” part of the market—the clean, semi-professional content—that is the highest quality user-generated or cheapest professional content. But to do so, YuMe, backed by Kholsa Ventures, Accel Partners and BV Capital, has opted for statistical data and metatags, which describe the content of a digital file, because cutting edge contextual technologies remain unproven.

“It’s a difficult thing to do, to analyze what someone is saying in English, deciding if the word was BMW, but also deciding if they said something nice about BMW or bad, before throwing a car ad up,” said Jayant Kadambi, CEO of YuMe. “Getting the semantics of that right on context is not that easy.”

Meanwhile, San Francisco-based VideoEgg, a specialist in videos on unpredictable social networking sites—backed by August Capital, First Round, and advertising giant WPP—uses computers and humans to vet 25,000 videos per day to make sure they are appropriate for the company’s 100 advertising clients.

Other companies have specially-named products geared to this problem, such as ScanScout’s Brand Protector, which CEO Doug McFarland says insures that advertisers will not have their ads placed with inappropriate content.

Indeed, with the transformation of entertainment into a world where homemade videos become pop culture icons and get distributed through computers, mobiles phones, and television sets, many of these companies are in it for the long haul. And if they succeed, they will have found the best ways to capitalize on user-generated videos while making advertisers, publishers, and users happy.

“I still think advertisers are wary of user-generated content,” said Andrew Frank, advertising analyst at Gartner. “It’s starting to change, but it’s a pretty long-range proposition.”