Breaking Uncanny Valley in CG Humans


Ideas to share the views of breaking the Uncanny valley in Cg.


??? :curious:


build a bridge?


They are getting pretty close with some new facial capture systems like the one
called contour reality capture
The average joe at this point doesn’t see much of a difference to be bothered by the uncanny valley effect.
However the rest of us do.

I still don’t see a point of performance capture other it being a tool out of many needed to create a movie. I’d rather see the actor himself on a CG environment, than a cg representation of the actor. Just use the cg version on the dangerous stunts.
I mean what do they hope to do with full cg actors, save money on wardrobe design?


“uncanny valley” was coined nearly 35 years ago by MASAHIRO MORI, Ph.D., a specialist in robot engineering and control engineering, and today it’s one of the commonly known design considerations of humanoid robots and synthetic characters.
Now where do we come in this?:slight_smile:
CG also we face the same problem of bringing human characters to life.
Questions asked ?

  1. Why realistic CG human characters lack the feeling of life?
  2. Is it because of animation flaws?
  3. Is it a lighting problem?
  4. Is in the area of effect/dynamics?
  5. Is it in the areas of facial animation?
  6. Is it in the areas of modeling?

Well thinking of this , why this is happening can’t we break this–
I’ll explain this in my following replies:)


Some sort of uncanny hammer.


Considering, that CG characters have been around for a very short time, we have seen a whole a lot of progress. IMO however several things need to be fully solved before we have any chance of reaching the UV. I wrote some down, which in my eyes still need improvement:

-Simulation of Facial muscles to achieve proper mimics
-Categorization of facial expressions to human emotions (see FACS)
-Categorization of human interaction both on a verbal and on a non-verbal level
-Physical correct motion of the human body (ever seen a mocap curve, it´s all jiggly and squirmish when compared to your normal animation curves)
-a proper SSS solver for skin simulation
-Simulation of inherant movement of the human body (heartbeat, breathing…)

If I remember some more, I´ll post them here.


How many thread like this do we need?


cuz people suck…


this has been discussed to hell and back. The problem is not that big: just stay away from photorealism when dealing with human characters.


Coffee came out my nose. :frowning:


I visited the Uncanny Valley once. Heard the fishing there was great, but I forgot to bring my real.


Ba dum chshhh!


Oh man… If I had the power to ban you, I would. :stuck_out_tongue:


You’re probably not the first forum leader to say that.


Ever heard of the search button? Regardless, I’m a researcher on the subject along with Karl MacDorman and you can read some of our results combining robotic and CG articles on the uncanny valley on wikipedia. Saying that, my personal opinion is that it will happen within 10 years (but frankly it’s an apparition of hope…shrug). You are already starting to see movement up the valley in stills and there was even more inching with very little facial movement like in the Ingrid Bergman piece. There are so many factors that go into the uncanny valley in CG that I suggest starting with the wikipedia. There are also quite a few academians who believe the uncanny valley doesn’t exist at all.


Hehe, sounds like that bridge to nowhere in Alaska.


[li]Use mocap in conjunction with hand animation for joints. Possible stereo based - to keep performance as real as possible through the camera - such as interaction etc.[/li][li]Track very detailed eye jitter/movement. [/li][li]Hand key facial performances off rotoscoped data frame by frame and exsentuate where neccesary, with nessecary facial targets 300+.[/li][li]Dont use humans.[/li][/ul]


December. Trailer was in front of the Crystal Skull, for those that saw it.