Video search application vendors are benefiting from the explosion of video
on the Web, from user-generated content on sites like YouTube to mainstream
media companies like CNN and ABC.
One recent newcomer to the market is Blinkx. The company powers video
search on Lycos.com and several other mass-market destination sites. It also
has relationships with approximately 100 different content providers,
allowing it to aggregate archival and fresh content from the likes of
YouTube, CNN and the BBC.
Blinkx.com also maintains its own site,
where it experiments with ideas, such as a video wall that visitors can use
to create their own content channels. Suranga Chandratillake, co-founder and CTO of Blinkx, sat down with internetnews.com to discuss the challenges facing his industry and his company.
Q: Video-search quality still seems far behind text search. Why is that?
It’s still definitely formative. [But] text search also took a long time to
perfect. It took a long time for Google to give you
that confidence that, if you do a search on Google and something doesn’t
come back, then it doesn’t exist. I don’t think video search is at that
level yet, in the same way that I don’t think that news search is at that
level yet.
The question is, is it just about finding a particular clip. How many times
do you want a specific video? And how many times do you actually want more
of a general content search? You want to know what’s happened on this topic
or this story. Has anything changed about this story since yesterday? That’s
really how we use video day today — we don’t use it for instructional stuff.
Q: What are some of the ways you’re addressing video-search quality?
We use speech recognition, for example, to literally listen to every
video, so we know word for word what’s going on inside. We’re also doing
video analysis with software that can basically watch the video and pull out
different scenes so that when we find something that interests you, we can
jump to that part of the video.
Q: How come you can’t simply search for someone’s face?
Actually, we can search for a face, but today we don’t use that function
because it’s not fast and robust enough. It will be there soon; it will be
our next big technological feature, but right now it’s not to the scale or
accuracy to get it right each time. I think it will be in the next 12
months; it’s one of those things where we’ll wait for it to be really good
before we launch it.
But we can do other things, like scene splits, by looking at how similar
each particular frame is compared to the last frame. Scene-change detection
is where you basically look at how much the scene changes from frame to
frame, and it helps you navigate bigger pieces of video more efficiently.
Q: Are you concerned that trademark and other legal issues will crimp the
viability of sites like YouTube or the online video market in general?
Even if that happens, what you get is eventually something like an iTunes.
If you think of a Napster, Napster failed as a company certainly. Napster
failed as a service. But it didn’t fail from the point of view of bringing
about a revolution in the way that we’re all able to access audio content.
ITunes is not free and not quite as pervasive as Napster was in its heyday,
but I can now very easily buy for a couple of bucks or a buck a single
track, and that’s a massive step.
People forget that for a long time, the
music industry was built entirely on the premise of selling two hits
surrounded by eight non-hits and making extra money from that, and that’s
been chipped away at completely.
In the same way, it’s almost irrelevant
what happens to YouTube. What really matters is that YouTube has proven that
people will watch video online even if the quality isn’t perfect and even
with ads all around. And all those things are important flags in the ground
in getting the traditional media companies, the people who own the content
we all want to watch, into that space.
It’s all pointing towards the YouTube revolution forcing the hand or
suggesting a hand to play for the big guys. Which is why we think we’re in a
great place. The whole premise of Blinkx is that this stuff will live in
lots of different places. You may not know where to go to watch what, but if
you know where Blinkx is, we’ll find it for you.
Q: What about bandwidth issues?
People are saying, “Oh, there’s more video online and that is going to make
the Internet slower.” Well, actually it turns out that for all the video
that is streamed on YouTube or CNN, there are many many more, much larger
amounts of data being transferred for uses that many of us aren’t that aware
of.
So for example, file-sharing sites and systems like BitTorrent, they
make up about 40 percent to 60 percent of all Internet traffic. There’s
another huge chunk, which is e-mail. E-mail is still by far the biggest
factor online and we all send many, many e-mails.
So yes, it’s true that there’s now more video and it’s true that the Web
portion of the traffic is growing bigger because video is part of that Web,
but it’s a small part of the overall pie. It will have an impact of course,
but it’s only one small part of what else is growing out there.
And as more and more video content becomes available online legally, and
also easily accessible, the amount of file sharing will increase at a slower
rate. Because if you look at what’s popular, particularly shows that come
out of the U.S. but aren’t shown abroad until a few months later, if ABC and
FOX start putting those online and available at the same time, then would
you go through all the hassle of setting up a BitTorrent client?
You don’t
mind watching a few ads, realistically, and you get to watch it full screen,
high quality and all the other stuff. So I think there will be an impact
where legitimate video impacts the growth of illegitimate video.
Q: The other question dogging online video is how do you turn a
profit?
There are lots of video sites out there. A whole bunch of them are paid for
by big media companies who make money from other businesses. So ABC.com is
paid for by ABC DVD or Disney DVD, let’s be honest. And the other half are
paid for by YouTube.
YouTube’s streaming bills are paid for because Google
has endless pockets. In that sense, everyone is struggling to find a model
for video that actually works, where revenue exceeds expenditure.
We think we’ve found it with search, because search is a very cheap thing to
do. When you click on a result, drawing in the result on the screen is
incredibly cheap; it’s just a bunch of XML. When you click on a result you
go to somebody else’s site and they’re the ones who pay for the storing
costs, the streaming costs and so on.
But we can put an ad right next to that experience, and so before you go to that site, you have every chance of clicking on that ad — whether it’s a video ad or a text ad — so it’s exactly the same way that people search; it’s a fantastically scalable mode for revenue generation.