Find all your DIY electronics in the MakerShed. 3D Printing, Kits, Arduino, Raspberry Pi, Books & more!

Bilde
This is amazing, on Tuesday Roger Ebert will be using his own voice (made from old recordings) while on the Oprah show.

After I lost my speaking voice, everybody thought they had this brilliant idea. “Hey! Why don’t you just take your voice from your old shows and put it on a computer?” Sounded good to me.
I kept getting suggestions: “I know this guy who says it would be easy.” Either there wasn’t a guy or he didn’t think it would be easy. In the meantime, I was using off-the-shelf computer voices on my laptop. My wife Chaz loved a voice named Lawrence, who had a British accent and sounded like a slightly crabby headmaster. Then I found a new Mac voice named Alex, who sounded like he knew when a sentence had ended. 

One day I was moseying around the Web and found the name of a company in Edinburgh named CereProc. They claimed they could build voices for specific customers. They had demos of the voices of George W. Bush and Arnold Schwarzenegger. (I amused myself by having them argue with each other.) In August 2009, I sent an e-mail to Scotland and heard back from Paul Welham, the president of CereProc, and Graham Leary, one of their programming geniuses.

They said they needed good quality audio to work with. Hey, no problem. I’d been doing movie reviews on television since 1975 and had hours and hours of old programs. But it wasn’t that simple. They listened to the old shows, and discovered (1) somebody else was always interrupting me, (2) I sounded all worked up a lot of the time, and (3) you could kinda hear the soundtracks of movies playing in the background.

….had an idea. Before I lost my voice due to cancer-related surgery, I’d recorded commentary tracks for some movies on DVD: “Citizen Kane,” “Casablanca,” “Floating Weeds,” “Dark City” and, ah, “Beyond the Valley of the Dolls.” These tracks had been recorded separately from the movies, so they could be edited to fit scenes. They might be “pure” audio. I asked two friends of mine, Ronnie Sass of Warner Bros. and Kim Hendrickson of the Criterion Collection, if they still had the original digital recordings. They rummaged in warehouses and found they did. So did New Line and 20th Century-Fox, studios for which I’d also recorded commentary tracks.

This began a back-and-forth process with CereProc, which had to transcribe every recording with perfect accuracy so they could locate every word. The “normal person” may use 5,000 words, not all of them on the same day. A college professor may use 15,000. Shakespeare used more than 25,000, but he was making up a lot of them as he went along.

Anyway, CereProc didn’t need to hear me speaking a specific word in order for my “voice” to say it. They needed lots of words to determine the general idea of how I might say a word. They transcribed and programmed and tweaked and fiddled, and early this February, sent me the files for a beta version of my voice. I played it for Chaz, and she said, yes, she could tell it was me. For one thing it knew exactly how I said “I.”

This was the voice I used in predicting the Oscar winners when Chaz and I taped a segment Friday of “The Oprah Winfrey Show.” When it was just me talking with Oprah, I used Alex. That show will air on Tuesday, so you can hear for yourself. Yes, “Roger Jr.” needs to be smoother in tone and steadier in pacing, but the little rascal is good. To hear him coming from my own computer made me ridiculously happy.

You can check out the CereVoice demos here.

Phillip Torrone

Editor at large – Make magazine. Creative director – Adafruit Industries, contributing editor – Popular Science. Previously: Founded – Hack-a-Day, how-to editor – Engadget, Director of product development – Fallon Worldwide, Technology Director – Braincraft.


Related
blog comments powered by Disqus

Featured Products from the MakerShed