Animator Getting Closer to Creating a Convincing Virtual Human Face

"DSC_19335", Image by Philippe Put

“DSC_19335”, Image by Philippe Put

[This post was originally uploaded on October 21, 2014. It has been updated below with new information on January 11, 2015.]

The one achievement that still eludes movie and gaming special effects artists and programmers is the creation of a human face so convincing that it could fool viewers into believing it is a real person. Vast and untold amounts of time, money and other resources have been expended in this quest and these artists and programmers have gotten close over the years. However, the human eye is so precise and discriminating that audiences can always accurately detect a virtual visage. In, well, effect, the imagery looks almost, but not quite “real”.

That is, maybe until now. According to a fascinating article on posted on October 14, 2014 entitled Have We Finally Conquered the Uncanny Valley? by Eric Molinsky, an animator named Chris Jones may have just achieved this or else come extremely close to it. He has been uploading his recent animation efforts to his blog, and two of them are also embedded in this article of a human face and a human hand. These are videos, not static images. I found the result to be extraordinary. I highly recommend clicking through to have a look at Jones’s efforts.

Let’s assume for a moment that Jones fully succeeds in his work and such virtual humans  start to populate movies, tv shows, videos and games. Lets further assume that the tools for doing this become widely accessible to computer generated imagery (CGI) artists. Then what? Here are my questions:

  • Whether and how will the careers of today’s real life working actors be impacted?
  • Will commercial audiences accept such virtual actors or will this be perceived as just being too creepy?
  • Will living actors or the estates of deceased actors be able to license their likenesses to be used in new video and film creative works?
  • Assuming that such licensing becomes a reality (even though the graphics remain unreal), what terms will be noticeable in terms of making an actor younger or older? What if the actor/licensor objects to final manner in which his or her image is used in the story?
  • Will new forms of agents and agencies be needed to handle the negotiations and contracts? Will future talent agents both literally and figuratively, become software agents?
  • Will new virtual and branded “stars” emerge in terms of the quality, usefulness and public acceptance of the imagery? That is, stars in terms of the virtual creations themselves and stars in terms of the CGI artists who emerge as the best in this specialty?
  • Will this development permit news forms of storytelling and gaming that are not possible with the current state of CGI?

I suppose then that this story also gives a whole new meaning to user inter-face development.

January 11, 2015 Update:

Breaching of the “uncanny valley” described in the original post above, the term used for the significant difficulty in creating a fully convincing computer animation of a human face, still remains quite elusive. According to a most interesting column in yesterday’s (January 10, 2015) edition of The Wall Street Journal entitled Why Digital-Movie Effects Still Can’t Do a Human Face by Alison Gopnik, this holy grail of CGI still presents some considerable roadblocks. I highly recommend clicking through and reading this piece in full. I will try to sum up, annotate and comment on it as a counterpoint to the post above.

Using a very clever analogy, the author compares the still unrealized feat of creation of a convincing CGI human to the Turing Test which, originally posited by the brilliant computer scientist Alan Turing (also the subject of the very well-received and possibly Oscar contending current bio-pic called The Imitation Game). This is a test for the achievement of actual machine “intelligence” whereby such a system cannot be detected in its interactions with an actual human being. That is, the human believes that he or she is communicating with another human when, in fact, the other party is a computer. Computer science is in deep pursuit of passing the Turing Test but it has thus far not been accomplished.

As between the passing the Turing Test and crossing the Uncanny Valley, Gopnik writes that the latter is “much, much hard for a computer to pass”.

While CGI is woven into so much of today’s visual media, they human face remains stalled in the Uncanny Valley for the time being. This is largely because of the incredible sensitivity of human vision and the wide range of subtleties in our human facial expressions to communicate our emotions to each other.

Gopnik further describes the effort on the fascinating ongoing project called Baby X by Mark Sager, a designer who worked the faces in Avatar and who is now a professor at the University of Auckland,  as being “one of the most convincingly complete computer-generated faces”. I highly recommend checking out and comparing this simulation to that of Chris Jones described and linked to above. Which do you think is more realistic and life-like in its appearance and movements? Does it make a considerable improvement in the realness of Sagar’s simulation that his project make additional use of the latest relevant neurological research when compared to the efforts of a highly skilled CGI artist alone?