OT: Need high quality Vocoder

farss wrote on 10/30/2004, 3:32 PM
I've done a bit of digging around and the only vocoders I can find are of the robot speak variety or else designed for low bandwidth comms.
Here's the project. Take sample of dead actor, have something analyse that and set parameters into vocoder. Feed new lines into vocoder to have dead talent say new line. Quality is important, speed isn't.
Maybe I'm being grossly optimistic but we seem to be getting close to doing this for images, speech would seem difficult but simpler than creating a 3D model of a face.
Bob.

Comments

MJhig wrote on 10/30/2004, 4:15 PM
I can't speak to the quality, haven't used it but I believe there's a free one at analogx.com. Use at your own risk ;-)

If you want I can post the link to "Cher's Believe" where they list the hardware used if you want to spend that kind of money and jump the necessary processing hurdles.

MJ
farss wrote on 10/30/2004, 9:19 PM
Thanx,
I found some sites with a fair bit of info, once I looked for 'speech synthesis' instead of 'vocoder'. There's some very serious work going on in this area and it does need a lot of manual intervention for things I hadn't thought of, the code isn't smart enough to work out the relative stress for each phoneme but doing a bit of manual work is OK.
Bob.
vanblah wrote on 11/1/2004, 7:47 AM
There was a project years ago to do this. I think it was Yamaha that was working on it. I haven't heard much about it lately.

They did a pretty good job of taking samples of voices and then having new lines being spoken by those voices.

If I can find a link to anything about it I will post it here.

EDIT - I can't find anything about this project. It had a name like Trillian (but that's not it).
vanblah wrote on 11/5/2004, 10:46 AM
I found it. It used to be called Miriam I think ... back when it was just in the development phase.

It's been released by Yamaha as Vocaloid:

http://www.zero-g.co.uk/index.cfm?articleid=804

They aren't planning on releasing the actual synthesis engine though ... just the singer libraries.

http://www.vocaloid.com/en/introduction.html
Doug