Breaking Uncanny Valley in CG Humans


But using sunglasses is cheating. Would it look as good without them?

I agree that the animation around the mouth was fantastic. I’m not so sure about the areas around the eyes, upper cheeks, and temples, although the sunglasses hide quite a bit of that.


Eyes are definitely a nightmare. My portfolio character proves the headache I’ve had with animating eyes. However, we know it is doable considering how incredible Davey Jones’s eyes were. Most people thought they were the actors eyes actually.


who have an article about the skin shader they used on MatriX and specialy how they made the animation around the mouth ?


I saw that presentation twice and if I remember correctly the animation was scanned by xyzrgb (thats a company name btw) EDIT: the actors were scanned, of course, not the animation : )


True enough, but Davey Jones also had an octopus for a face. That makes any question of the uncanny valley kinda moot…


Acting has nothing to do with it. Acting and storyline are only important for characters who are incomplete (cartoon) because you have to be entertained to empathize with them.

How can acting have nothing to do with a character giving a performance? Acting is the essence of the thing.

What we call the ‘uncanny valley’ effect is an emotional response to a performance given by a synthetic character on screen, a performance that falls short in way that is hard to explicitly define- hence the use of the shorthand term ‘uncanny valley.’

We take it for granted when we watch a good human actor on the screen that we can ‘read’ his characters inner states- we seem to have access to his thoughts and feelings. But this effect is not accidental- it is what screen acting is all about, and is a style of behavior not seen in everday life.

So in my opionion the quest to put a computer generated ‘real person’ on the screen is a flawed project. What needs to be replicated is the art that allows a good screen actor to give the audience an inside view of the characters they play.

Consider this; when you are watching a drama and a moment comes when you realise that one character is lying to another, and you think to yourself ‘ah ha- I can see he’s not telling the truth’ why is it that the other characters in the scene cannot see the same?

You are being cued in by the performer that his character is lying to the other characters, cued in by subtle gestures and facial movements that by convention the other characters in the scene cannot see.

So there are framing rules in play here that allow all sorts of things to be achived that are nothing to do with ‘real life’ and everything to do with dramatic staging and craft of screen acting.

Without a deep understanding of these arts and the framing conventions that govern them, even a pore perfect CG human will come across on screen as dead as last weeks haddock.


Acting is a vanier. Accepting or failing to detect an organic recognisable artificial phenomenon has to do with eating and screwing. The movements belong to your aunty and the baker. You must eat and reproduce so your primative talent enables you to pick uncharacteristic movement in a herd from a great distance. That is the one you go after because it is easier to kill. In a bar you can see which animals are ready to mate.

Shake someones hand for the first time and apart from all the other synaptic information like temperature, pressure and odour there is a load of information you process instantly, will the aquaintance make a good hunting buddy, do you trust them and so very much more. If you shift your weight and billions of fibres, tendons, your skin slides, I can see if you are sick, fit or an electric donkey bottom kisser.

At the base level it is the trillions of alterations per second everyone makes that is the determining factor apart from outward appearance. These are made by all people not just actors:cool:. The next level is gesture like women running their hands through their hair (preening) for this read the Naked Ape by Desmond Morris. Your sensory perception is so good you can see a lighted match from a mile distance in total darkness, detect taste of one grain of sugar in a whole barrel of water. If we get the base perception right we can move onto the other stuff.

Then like someone said we do actually have what I like to loosly refer to as a soul. This has nothing to do with actors but rather survial tools we have and use so very well everyday.



The eyes were not animated and the octo was matched over a naturally moving physognomy which was in turn attatched to a naturally moving body. Comparison cant be made. Even so you dont see too many people with a face like that for primeval referance, that goes for the animated skeletons as well. The creepy stuff begins when you see a human you are used to encountering in your daily routine. The question at hand remains very much an issue.

Just because it is fun to solve!


Acting is a vanier

I understand what your saying, but what you seem to believe is that to solve the ‘uncanny valley’ issue you need to reproduce a real human being on the screen. This is false problem.

The figures you see in movies are not behaving in a naturalistic way- real people do not signal their intent and emotions in the blatent way way that screen people do. They seem real enough in the context of the movie, but this is a special context.

Good animators understand intuitively that in order for their characters to appear ‘real’ they have to ‘open them up’ so that their inner thoughts and feelings are easily read by the audience (but by convention not read by the other characters, unless this is part of the story.)

The way this ‘opening up’ is achived is by grossly exaggerating the verbal/nonverbal cues that humans use to understand each other.

The fact that we readily accept these characters as ‘real’ despite their grossly exaggerated displays of intent and emotion is because humans seem to have an innate tendancy to see the world in terms of theatrical forms- we are a narrative driven species- most of our talk takes the form of stories- he said, she said ect.

Consider the degree of reframing required of the average person watching a movie. They enter a world were viewpoints move instantly in both space and time, where perspectives shift from close up to panoramic at the drop of a hat, or zoom rapidly from one point to another. None of this bothers your average joe at the movies- he takes it in his stride- despite the fact that if the real world behaved like that he would freak out. So as a species humans seem to have a remarkable ability to adapt to radicly different frames of reference when it comes to narrative experience.

So a convincing human screen character is not the same thing as a real person. Uncanny valley is, in essence, a failure of staging, not a failure to replicate a real human being.


The effect is not confined or indeed only usefull in movies. As for many years to come this will be your only frame of referance I can hardly blame you for hammering on this 2d medium.
People generally relate better to animated figures that are distinctly outlandish than those that begin to approach the ideal. This is a phenomenon known to robotics researchers as "the uncanny valley"—that point where a robot is so close to lifelike yet still so short of ideal that people become focused on its imperfections.

[i]Feelings of unease, fear, or revulsion created by a robot or robotic device that appears to be, but is not quite, human-like.

[/i]Automotons will have real commercial applications aside from the entertainment industry in the future. Already they are being used (experimentaly) as nurses for the elderly in Japan. The effect we are refering to means you will not wait for a character to tell you a ripping yarn before your skin begins to crawl. A good animator and a good studio like Pixar you mentioned are painfully aware of this problem. Funnily enough you only refered to the toy characters with nostalgia but not the human cartoon characters in toy story. That should tell you something. You must know the reason for the legions of fluffy character productions is directly related to this. As yet no ammount of smoke and mirrors on the screen has overcome the puzzel.

Your suggestion that narrative is the solution greatly oversimplifies the entire fenomenon.


This is why I hated those “human” puppet shows like Thunderbirds, they creep me out.


I am not really understanding why you guys are talking about animation when we’re obtaining quite impressive results with Mova and Image Metrics or even Face Robot. At that point it’s more a matter of acting and rigging. Keyframing facial animations is not really something I wish to try out though, I’m going to stick to Compositing and Modeling…:argh:

What I don’t like about actual CG characters is that we do not see the inside of the mouth enough, these teeth on the Agent Smith are impressive. I don’t like the hair rendering too, it’s looking too soft and blurry. Bah, I think for still images we have to improve the textures resolution and the materials and had some subtle hair, even girls have facial hair…

Davy Jones is my favorite one so far. I know it’s an octopus but the animation and the rendering is superb; the SSS, the displacement maps, the flawless lighting and integration in the environnement are enough reasons for me to praise him as my favorite CG character.

Avatar probably is going to be the next big thing… I heard they’re making many CG characters and that Weta’s on it.


to improve uncanny valley we need to start building some residential homes on that area and some 7-Eleven.


Of course the eyes were animated, all of Davy Jones was animated.


Ah it seems you are right, the director asked for eye shots to blend over which they did (hence the makeup). Seems they never used the actual footage. In that respect I stand corrected. However it wasnt my point.

People generally relate better to animated figures that are distinctly outlandish than those that begin to approach the ideal.

This was my point. I am not quoting the article because it is right but to my mind a logical explaination. You stated in an earlier post that the valley was moot while monsters and cartoons are what we call falling into the non idetification range. It’s not an argument, its a problem and a challenge. It’s one that is being worked on very hard because the advantages are numerous, especially in robotics. Also we are learning alot about our perception. For people engaged in synthetics I would have thought we coul;d see that. Visit the link I provided in an earlier post maybe.


The effect is not confined or indeed only usefull in movies.

No but this thread was directly related to CG Humans. I agree that robots are a different thing entirely.

I suspect that if humanoid robots are ever widely used in roles that involve interacting with people they will be designed to look as ‘cartoonish’ as possible to reduce any sense of threat.

I don’t see any real advantage in robots that try to look like people.


Davey Jones and Gollem are examples of where the uncanny valley effect is actually beneficial to the result, as the extra element of creepiness helps such characters.


If I remember correctly from an interview on the extras from the toy story dvd, someone from pixar talked about purposeley making the dolls seem more real than the humans. So they used the uncanny valley to their advantage for story effect.


Okay, I am not going to check back and see who wrote what, as this thread has gotten rather long, but rather just address some of the general points.

The uncanny valley exists, and will continue to be a problem until advances are made to overcome it. I also think that the advances will be a matter of TECHNIQUE rather than TECHNOLOGY. CG is an art form, and requires artists. Consider that a life cast of a human face often looks fake, while a sculpted one seems real. Let’s take a look at both these cases:

The fake-looking cast results from the subject reacting to the process of making the cast. Actors standing perfectly still while their faces are scanned by lasers are going to have a hard time keeping even a relaxed expression. Their acting will also be hampered by the unfamiliarity of tight, itchy mocap suits, or little reflectors stuck all over their faces. There are muscles in the face that very few of even the best actors have complete conscious control over. This, along with other related phenomena, explains the effectiveness of method acting. Actors working on a green screen often come across just as fake and stilted as CG humans. When actors first began to make the transition from live performance to film, they needed to unlearn a lot what they knew. Motion capture is another kind of performance yet, and we are starting to see specialists like Andy Serkis who do it particularly well. I think an actor looking to get into motion capture should try wearing the suits around all the time, until it feels natural. Otherwise, the capture data will reflect their unease in the suits.

The sculpture that looks real is made by a person who understands both the minutiae and the entirety of the human subject. I think the problem with a lot of inhabitants of the uncanny valley is that their creators were too focused on the details, and forgot to think about the whole. The benefit of exaggerated or fantastic characters is that the animators are freed to think about the characters in broader terms, and how the parts relate to the whole. When your goal is entirely to make a character photorealistic, it is easy to become focused on doing just that, and lose perspective of the character’s personality and place in the story. Sometimes a step back really helps, and when an FX company is behind schedule or over budget, that step back is the first thing to go.

One of the biggest problems is eyes. We are born with the instinct to respond to faces, and the eyes especially become a center of focus. Even cameras can recognize faces now! Some of the the common problems I have seen is eyes that are either too shiny or too dull, or are either too wide eyes or too sleepy looking (we are talking about a feature that can convey completely different expressions with changes of less than a millimeter). In real life, eyelids move up and down as the eyes look in different directions, and the lower lid pulls toward the center of the face when the eyes close. These are both things a lot of animators get wrong. The trickiest thing, I think, is that when we look around we use a combination of moving both our heads and our eyes. I’ve no idea if there are any “rules” for how much the head moves versus the eyes when we look around. It would be helpful to know.

CG really has come a long way in a short time. Think how many lifetimes it took the human race to progress from stick figures on cave walls to the ceiling of the Sistine Chapel. Looking at the advances made in our industry in just a few decades is really impressive. And unlike other advances in computer related fields, it is not entirely a matter of bigger and better software. As individuals and as a community, we artists are continuously pushing the boundaries. The uncanny valley is just one of those boundaries, and when it is crossed I don’t think it will be a result of a new technology, but of someone finding a better way to use the technology we already have.


This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.