Software That Morphs A Face For Talking?

Hey all, is there known software that, if you give it a frontal view of a face, can morph it intelligently to match the lip movements of speech? Even if I had to tell it where the facial features were, and specifically what word was being spoken, that would be great. I know there’s a mod for Skype called Crazytalk that does something like this. And the site Gizmoz has a primitive form of it, but I was wondering if there is some standalone package that would allow a graphics dunce like myself to pull it off? I want to animate the face of a toy dinosaur so I can make it talk.


Do you mean you have a 3d CG dinosaur, or an actual physical toy dinosaur?

The difference between animating the two of them is very great.

We use servo motors on anything physical we need to talk, but the more realistic you want it to be, the more servos it requires, as each little facial movement requires another servo.
These are controlled by Gilderfluke software.
Lots of $$$$$

But I suspect you want something cg…in fact something like the Genius Webcam software that recognizes your face and then morphs you into an alien or such, then copies what you do.
Sorry no other ideas on that.

>> Genius Webcam software

That’s more like it. I have robots but I want this to be animated. Do you have a link to any sample video captures made with it? I Googled “Genius Webcam” and got the site, but didn’t see any sample videos.

Also, do you know of an outlet that sells them to the USA? The stores I found, like Kelkoo Computing, were UK based.

Thanks again.

Actually :o… You can do that with Blender. Unless you DO mean a physical object.

Spins’ Post on the Meloware lipsync script.
Magpie & python script combination.

The original script is here

Correct me if I’m wrong but those script solutions still require you to make the phoneme shapes first - don’t they? And really, once, you’ve done that you could easily do the lip sync manually in Blender.

If there’s software that does exactly what roschler has asked for I’d be concerned about the quality of the output and I assume it would require a fairly specific loop/mesh structure in order to work its magic.