OK, so here’s a thing. So apparently Youtube has an automatic transcript function, where it will use voice recognition to try and build a transcript of your video. You might think this presents all sorts of interesting IP issues, and you’d be right. Whatever, not relevant right now. The important thing is that their voice recognition technology SUCKS. I mean, beyond the earliest betas of Babelfish bad. Beyond the most hilarious coffee table book of Engrish photos bad. No, let me tell you how bad it is. It is so bad, that it no joke creates rap lyrics that are better than 99% of what is out there today. For example, out of this video (a commentator doing play-by-play of an online game tournament) we get the following:
“22:48 affleck in which the person’s trying to sneak it thus delta territory
22:52 and its most places where it’s kind of imbalanced
22:58 groping for missile dean it appears that yet but i think this withholding the irs”
Or this one:
“19:06 based on the thanks mrs unearthing the control
19:09 even that could have done earlier should’ve been under earlier
19:12 wasn’t done earlier then you need to
19:14 really added into your falklands
19:18 along israel’s i’m kathleen turner planning
19:22 wide web points watson position for what
19:26 where do i have things
”
You can read the whole thing at the link below — just beneath the video, look for what looks like an icon of a piece of paper going through a fax machine, if it is there (may not be enabled on all videos) and select English (automatic captions).
Guest post contributed by Dan Rosenthal
One Response to Guest post: YouTube’s Automatic Transcript Function