Ho John Lee’s Weblog

Been thinking a bit lately about dealing with video, converged media, and search, came across a couple of interesting pieces on video search and digital content in general, first one on John Batelle’s SearchBlog, which in turn references a longer article by Mark Glaser at the Annenberg Online Journalism Review.

Ourmedia, SingingFish, and Brightcove are profiled briefly, along with Google Video Upload, Yahoo Video, and Open Media Network.

From the Glaser article:

Howe estimates there could be 300 million video streams online, but Singingfish has still only scratched the surface with just under 20 million streams indexed. Singingfish also crawls adult content — literally anything that’s legal — and includes a “Family Filter” with pretty conservative rules for what partner sites or individuals can filter out (including sex education material).

Finally, Howe believes that there’s been a sea change at media companies when it comes to embracing video search. “There’s been a general recognition that they’re going to have to digitize their content, and if they’re going to digitize it, then they’re going to have to monetize it,” she said. “I think people have sort of gotten over themselves. They used to assume that people would just go to such-and-such site to find this wonderful content. Well, no, because people have so many options.”

Some related thoughts:

The metadata problem:
It’s going to be difficult to make Google-style searches for video work without lots of Flickr- or Technorati-style tagging and other metadata sources. This can work for communities of motivated individuals sharing an interest in a topic or body of work, or for a successful commercial movie or television program. But for hours upon hours of uncut home videos from DV camcorders or the more recent digicams, even the owner probably won’t have the time or interest to tag the content enough to make it usefully searchable for most applications. Another problem — people generally have to see the video content to apply tagging or other metadata — and it’s just hard to get the data there…

The distribution problem:
Today’s internet isn’t well suited to moving large data files to and from end user sites (i.e. homes and most businesses). Relatively popular content can make use of peer-to-peer technology like BitTorrent to recruit end users to redistribute the original content and spreading out the bandwidth demand to other parts of the internet, but this only works if clients participate and if the content is popular enough to develop a network of clients with cached copies. Kontiki appears to be building a different peer-to-peer content distribution system.

Commercial sites sometimes use content delivery networks such as Akamai / Speedera (currently in the process of merging) to move copies of media data such as video, audio, imagery, or multimedia (usually Flash or Java) closer to the expected network clients. The underlying source of the infrastructure funding is often the online advertising dollars spent in marketing campaigns by movie, television, music, automobile, and lifestyle products. This doesn’t mesh well with a grassroots model.

If grassroots video is to become widely used, it needs to become accessible. On a good day, the DSL line to my home manages 1.5mbps down and 384kbps up. If I want to share a few minute digital video clip from the DV camera, it could easily be hundreds of megabytes, requiring more than an hour to upload to either a peer-to-peer network or a content server. I could also recode the video to make it smaller, which is a common practice for video publishers today, but this requires knowledge, tools, and time.

Community tagging and metadata for video?:
Sites like Flickr allow a community of interest to build around tags representing common interests, and also allow a vocabulary of tags to evolve, along with a social network of people who find each others’ photos interesting. However, this presumes that people can actually see the content they’re tagging, which may be difficult for a while in the case of video. I’m assuming that Google Video Upload (and others) will probably do some basic tasks such as segmenting on scene changes, timecode breaks, and perhaps simple scene analysis. But without anything else to work with, search engines aren’t going to help much.

Enough for now…

This entry was posted on Wednesday, April 27th, 2005 at 11:15 pm and is filed under Media / MP3 / PVR, Business, The Internet, Search Engines. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.