Searching *inside* your photos, podcasts, movies and more…

When I was at Vignette during and after the dot.com adventure, we were the web content management engine behind numerous marquee sites with rich media.  One of our partners — on whom I drank the KoolAid about their value proposition — was a company called Virage. These guys had an unbelievably cool technology that could essentially index a *huge* set of data about what lived inside unstructured media content: facial recognition; the closed captioning track; real-time analysis and encoding of streaming media (which to me was THE COOLEST thing); and a whole lot more.

Virage’s customer list is a who’s who of media companies globally. In the summer of 2003 they were acquired by Autonomy who has also sold licenses of their analytics for the U.S. intelligence services (seemed like a good fit: the Virage sweet-spot is media…Autonomy’s is static content plus context and more).

I did a little “dot connecting” this evening. 

Tonight I was reading Doc Searls blog entry about frustration with tagging his photos so they can be found. Later I was looking at some video content on a blog (I’ve been exploring video blogging or “vlogging” lately) and also organizing the podcasts that were downloaded today to my machine. The groundswell of rich media content creation is palpable and I think it’s pretty exciting — if you can find the great content being created.

Instead of just stumbling across this content while poking around (except for the podcasts which I’ve made a point of selecting), what if I was actually looking for something specific? What if I’d wanted to search for a particular scene of a video or for snippets contained within a podcast? This is currently a really hard problem to solve unless some human has input text (like closed captioning for video) or built metadata around an audio file — but is really tough to do on-the-fly and automatically.

Virage has done some interesting things (like facial recognition) but I’m not certain where the state-of-the-art is with accuracy. I recalled Virage while I was thinking about this problem tonight (searching inside and for rich media content) and wondered what engines for media metadata creation might look like, what the market opportunity might be and what “dots” I should connect:

• My own post about drowning in content and, frankly, even finding it,
• Concern by people like Doc Searls over how to tag his own photos (which seems significantly more simple to solve than creating searchable rich media),
• The acceleration in content repositories like the Internet Archive,
• The proliferation of do-it-yourself tools like iMovie, Garageband, Audacity, Premiere, which ONLY means that we’re going to have even MORE rich media we’ll want to find
• Continuing reduction in price and increase in power of PC’s and video/audio tools,
• Blogging offerings like Audioblog that stream media to-and-through blogs
• Virage/Autonomy could solve a lot of the above as a hosted service.

There is no question that there is a coming tidal wave of unstructured media that is going to be VERY hard to find — unless the right server-centric horsepower is there for all to use. I’m gonna take a guess that someone, somewhere is or has been chatting with Autonomy and Virage about just such a service to compete with the gorillas in search.

Of course, our search gorilla pals at Google undoubtedly see this clearly and have kickstarted a beta search offering for video. I have a sneaking suspicion that they see this problem/opportunity. I also suspect that they understand the blogging demand that is growing (since they acquired Blogger) and might very well be working on this problem on a large scale.