Maybe this will work out fine, but I worry that with a video you'll pick up things like "this guy seems uncomfortable" but not be able to identify the reasons why (is he uncomfortable with the rest of his group? Or just uncomfortable being in front of a camera?) -- or worse, get a subjective sense that a video isn't very inspiring without even identifying what you're noticing, never mind figuring out whether what you're noticing is signal or noise.
There's a reason why most professional symphony orchestras use blind auditions: While nobody disputes the fact that being able to see a musician play is valuable, experience has shown that the noise it introduces is more significant than the signal. Simply put, eliminating the high bandwidth distraction of being able to see someone allows orchestras to pick better people.
Higher bandwidth indeed, but you're giving out an open canvas. Could be good, could be extreme. The restrictions of "traditional" submission mechanisms are like the "traditional" formats for resumes, preventing every resume from becoming a tabloid magazine.
http://ycombinator.com/video.html
If they ignore us and put in a bunch of fancy stuff anyway, we'll just click on "next." We have a lot of applications to read.
But people shouldn't worry that they "video badly." We're not (I hope) so dumb that we'd be taken in by people who are merely smooth presenters. We come from the world of people who video badly ourselves. If we'd been applying for Viaweb, I'd have had to negotiate with Rtm how many words he'd say, and I'd be lucky if I got it as high as 10.