There is a concept in artificial intelligence called "the uncanny valley." The idea is that somewhere between going from robot to human, there is a valley where "almost human" appears strange. You have a negative emotional response to such a creature.
[Uncanny Valley.png]
Things like robots and stuffed animals and cartoons look clearly
not human, while some animation (like, for example, in the
movie Polar Express), prosthetics, and the animatronics you see at Disneyland look
kind of human. That grey area, where something goes from not quite human to human, is the uncanny valley. CrazyTalk, an application that will sync photos with voices, creates videos that sit in the uncanny valley. They are fun to make, for sure, but also somewhat weird.
Once you get started with CrazyTalk a wizard walks you through selecting a photo and then choosing key points on the face.
[wizard1.jpg]
One cool thing is that the software is pretty good at finding where the mouth and eyes are in an image. I hardly had to move the dots at all.
[wizard2.jpg]
Once you have your photo loaded, you can use the built-in mask editor to clear out the background and, if you want, put in your own background. This is a feature I wish more multimedia software included.
[mask.jpg]
The final step is to edit the script. In CrazyTalk, a script is the list of audio and facial expressions that go along with your animated photo. You can either put in text-to-audio or record your own audio. At key points, you can choose facial expressions and CrazyTalk will animate the face.
[audio.jpg]
That's pretty much it. The
realness of your animation comes down to how much time you put into scripting. You can also do some really advanced work with the eyes and mouth. For example, when the photo talks, do you want the teeth to display? What color should the back of the mouth be? All of these are editable in CrazyTalk.
[advanced face.jpg]
