The Boston Globe has an interesting article on one of the projects out of MIT's Center for Biological and Computational Learning (At MIT, they can put words in our mouths) [via Metafilter]. MIT researchers have developed software that permits the creation of extremely realistic video that shows people saying things that they have never said. The article includes four sample videos, which are incredible (although low-resolution web-streaming quality). While the process has its limitations (only works for one or two sentences, subject must be facing the camera as a news anchor would), the potential of this is enormous. CNN will love this. If more fully developed, there would be no more need for foreign anchors, just more computers in Atlanta. Movies could do away with cheap dubbing - foreign actors could be made to appear to mouth English words (or vice versa). Of course, blurring the reality of video has tremendous potential for fraud and will certainly raise evidentiary questions (though most such objections for the forseeable future will be for the novelty value). More important and immediate I think, though, will be the legal issues involving copyrights, privacy rights and rights of publicity (start modifying those artist's contracts now).
In related news, WIRED Magazine has a lyrical article on the difficuties and triumphs of synthespian animation (Why Is This Man Smiling?). Realistic facial animation may be sometime away, but progress is steady. Certainly state-of-the-art rendering is capable of realistic humanoid, if not human, character animation. For more on virtual reality, see also LawMeme's (The Future of Virtual Kiddie Pr0n and Other Notes on Ashcroft v. Free Speech).