May 31, 2008


Flash video to text conversion

YouTube brought videos to internet. Really. The idea of making webpages embed the video provided instant gratification which was missing in "Click here to download the video in wmv format." A whole lot of interesting content is now available via this medium. But it has created a new problem. In the good ol' days, the transcripts of interviews at conferences were sincerely prepared and put on web. The videos have made these people lazy. Why prepare transcript when people can actually see the live people talking? Well, for one, I can read a whole lot faster. Secondly, I really don't need to associate faces with names. So here is my tiny idea to address this problem.

The service will sport a google like text box with a button titled "Get the text." You can copy-paste the URL the page on which video is embedded. Once you click on "Get the text" button, it will ask you take a chai break. Meanwhile, the service will get the page, fetch the video on the page, extract the audio, then use some speech-to-text converter to prepare the text. It can also label the voices as "person 1", "person 2", etc. Once you are back, you can specify the actual names for "person 1", "person 2," etc. Bingo! You have the interview transcript ready.

As Scott Adams says, you will either find this idea impossible to implement or you will point out that somebody has already done it.

It's a great idea, would love to see it. Unfortunately, that kind of speech-to-text technology just doesn't exist and may never. The best programs (Dragon Naturally Speaking, etc.) don't do well with multiple speakers, unknown voices (you have to train your voice first in some versions) and with technical terms. Human transcription just works better, with fewer mistakes. Course, the rates for that (we charge $1 per audio minute) are not affordable for everyone. But maybe if some advertising were integrated with it...
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?