Breaking Uncanny Valley in CG Humans


#21

lol, totally agree


#22

Well What you said is right

In fact all these points are important criteria for bringing characters to life, but not this alone .Psychologist say, your subconscious knows what real humans behave like. We spend our lives reading unspoken messages from others, even though we’re not fully aware of it. The missing of subliminal expression was main draw back in ‘Final fantasy’ and ‘Polar express’ etc. In an article I read that in Polar express they even added breathing rhythm animation in most of the static animation shots.

[b]Psychologist[/b] like Peter Plantec and Paul Ekman studied Human behaviors for decades. I’ve been reading FACS by Paul Ekman, these technique is widely used in leading animation companies in the industry.

  Well this alone doesn’t solve the problem of uncanny valley in Cg but lighting, hair/cloth simulations also contribute this to an extent. If facial expression and animation contribute 60% rest of the elements contribute 40%.

Even modeling takes important role especially in areas of eye and lips.

     Pure data capturing using our technology alone doesn’t provide solution ,it’s also a psychological feeling related to our subconscious mind.


#23

One thing that always surprises me is how often I can look at something and study it carefully over and over and not see it. Then one day bam, it’s there! I see it. Once I finally see it it becomes more or less clear sailing to draw it, render it, shade it, or whatever. About the time I start to think I’ve cleared some milestone of understanding it happens again with something new. Has this ever happened to you?

Anyway, my calculation is that this is basically an infinitely deep rabbit hole and one that makes continually striving to be a better artist so rewarding and worthwhile on its own.

My theory then is that no one person or group or persons is ever going to have the initial requirement of being able to “see” enough of what makes something human to be able to translate the full range of human behavior into an animated 3D character that can fool many people. Progress toward that end will always be met with moving goal posts just that much more out of reach.

In my opinion it can’t ever be achieved, and if I ever am fooled into thinking a 3D character is human then I think that would not reflect well on my own powers of observation. As the previous poster alluded to, it’s much easier to feel something missing than to identify what should be there.


#24

That’s pretty much my take on this matter as well, if we exclude all those really fast moving fraction-of-a-second type VFX shots with digital doubles.

Additionally, another factor to consider is that real humans tend to be depicted in exceedingly “synthetic” ways in movies and TV shows - this, I believe, makes people forget what humans naturally look like in the context of entertainment. Everything is very carefully stylized with lighting, make-up, retouching, grading, tweaking, editing… and ironically enough this brings reality closer to the synthetic, instead of the other way around, which in turn can make you believe that it will indeed be possible to create a CG actor capable of conveying subtlety, emotion and life indiscernible from the real thing, when in fact it’s a case of the original starting to emulate the facsimile instead.

It’s when you instead look at very casual and/or low budget productions that the insurmountable difficulties become much clearer. Take any random episode of those endlessly crappy “lets build a chopper” documentaries on Discovery channel, or some really low budget feature film like “Pi” or “Primer” - Look closely at the people/actors as they’re just doing whatever it is they’re doing… No matter what amazing future technology you’d have at your disposal, there is no way you could replicate even five minutes of that stuff and have an audience believe they’re looking at real humans.

So, can it be done in a very tight, brief couple of seconds flashing by when intercut between several minutes of footage of the real original human actor? Yes, absolutely, it already happens on a regular basis.

Can it be done in terms of anything that a real actor can do, for several minutes, or even 90 minutes, without looking wrong? No, I really don’t think so.

.


#25

Uncanny Valley is only a phrase to describe something uncompleted.

It means that we are trying to do something complex that we can’t even understand. In future, we will remember this with fun because Uncanny Valley has no real meaning. It is not a discovery.

The same analogy applies to modeling/topology. simple turns better because it is easier to control. But something complex, requires complex topology, which requires complex rigging.

Movies like Beowulf, is a clear example of the effort to hide the lack of detail, depending too much on textures instead of motion and emotions.

Breaking the uncanny valley? sure, it’s simple. study, research, become a mad man.

·


#26

The problem is that it would take someone that really was truly mad… as in “clinically insane”… because, if successful, you would have spent years or decades to accomplish something that could easily be done in three minutes just by sticking a camera in someones face and press record and say “action”.


#27

As a non-animator (occasional dabbler), I think there’s a long way to go to transcend the Uncanny Valley effect and achieve the Holly Grail of convincing human simulation in 4d. Most realistic simulations of Humans still don’t express a convincing compliance with Newton’s 3 basic laws of motion, let alone any reasonable sense of personality to fool me. The best sims still tend to look environmentally disconnected, massless and mechanical (lacking aesthetic grace in movement where needed). Poor hair and cloth simulation sure doesn’t help either. The mind may be easily tricked but in the scenario of human simulation in animation, it is exceptionally hard to fool.
—My 1.28¢ (in euros)


#28

Ingrid Bergman piece is a nice one i’ve seen it too. As you said in CG we already achieved it in still frames but when it comes to animation it’s started showing.

I believe if we pay attention to our subconscious feeling in every frame of animation and lighting that we could achieve in every shot that is required. Frankly it’s impossible to sort out each subliminal expression and light falls on a character but a stage will come that our own eyes will be convinced like this still image of ‘Ingrid Bergman’.

  There are so many factors that contribute for uncanny valley but what contributed for that believability of the still image ‘Ingrid Bergman’ is a interesting area to look into. If we look into the same lines in consecutive frames of animation then, this brings the characters to life.

In my opinion it will take another couple of years to see believable CG human character in screen. Hope avatar will set a trend for that.

  ------

#29

Why bother? Stylized characters are like teddy bears, they are lovable and cute dispite what’s missing. Emotion and acting show through much better with stylized characters… so why would you want to make a super realistic human and use it for anything more than stunt doubling and video games?

The challenge? Maybe, but the audience doesn’t care how good you are. They want to be entertained.

Just my 2 cents.


#30

Oh my god, that was hillarious. What is that from?


#31
 Because we as humans have always been fascinated with the human form. Why do we still do life drawings or sculpture for that matter, other than to improve our understanding of the human anatomy and it's proportions? As well, we want to tell stories that are not always constrained to the stylistic and cartoony realm. Stories such as Beowolf and the Odyssey (or even a flying human for a marvel movie, etc.) simply are too much trouble to use just a human actor in some respects, because of the fantastical world involved. Regardless, the uncanny valley was more originated for robots, because scientists are trying to understand whether a realistic robot of the future would be more accepted by the public than say Asimo. [http://www.honda.com/asimo/?ef_id=1097:3:s_f027e43c2c8f8f1454f1945123c723d8_841609828:2NoZxtB6B3YAAAcJROgAAAAF:20080531182237](http://www.honda.com/asimo/?ef_id=1097:3:s_f027e43c2c8f8f1454f1945123c723d8_841609828:2NoZxtB6B3YAAAcJROgAAAAF:20080531182237)

And if you believe what the uncanny valley is cooking, then probably not.


#32

That’s from 30 Rock, the Tinay Fey show on NBC.


#33

Mr Positive, I always thought we did life drawings to improve our ability to draw characters of all types. No matter what type of character you’re looking to illustrate, no matter what style you’re employing, knowledge of basic anatomy can help a great deal.

Besides, it’s one thing to draw a realistic human. It’s a whole other thing to model, texture, light, animate, and render one that looks real. At a certain point you just have to ask yourself - would it be better to just use a real person?

I was a little disturbed with the characters in Polar Express. It distracted me from the story. On the other hand, the deformed cartoony characters in Ratatoli were absolutely perfect, and I was able to get into the story.

Which one is more likable? A mannequin wearing cowboy clothes - or Woody from Toystory?


#34

Fun fact: “uncanny” was used to describe 2D “Snow White” when it was first released.

In some of the early sequences there may be an uncertainty of line, a jerkiness in the movements of the Princess; but it is corrected later and hand and lip movements assume an uncanny reality.

  • Frank S. Nugent in The New York Times, 1937

#35

Why not just use Python?

[b]import [/b]soul

(obscure reference for the xkcd fans out there)


#36

**that fact was indeed fun.


#37

Because we have to learn to see. You will distort what you see anyway, and when the image from your mind gets to the hands is distorted again. The point is: to get the image out of your head as clear as possible: you need a steady hand for this.
This distortion (usually) is ok in the end, basically you emphasize the personality.
Probably this is why art teachers encourage you to exagerate a little.


#38
  Seriously, I'm not exactly sure what your point is. You have described the phenomena of uncanny valley. Shrug. Regardless, if you don't believe that creating realistic humans is not of the utmost of highest endeavors, regardless of the medium, then I suggest you check out the top viewed cg stills of this site.
  [http://forums.cgsociety.org/forumdisplay.php?f=121&daysprune=-1&order=desc&sort=views](http://forums.cgsociety.org/forumdisplay.php?f=121&daysprune=-1&order=desc&sort=views)
  Let's face it, people are subconsciously interested in the human form in art and especially realism (and yes stylism as well....it's the in between that is in question). As well, what you mentioned for drawing and needing to understand the human anatomy to draw all types of characters, should not be just excluded to 2D, and should easily parlay to 3D as well. Again, I'm not exactly sure if we are disagreeing on anything.

#39

Yes, subconsciously our mind likes to see realistic images of characters


#40

I believe it goes back to what the grand master of make-up artists Dick Smith termed the Dr. Frankenstein concept. Of course he was speaking of foam latex make-up, but there comes a time during the process when something you’ve created begins to take on a “life” of its own.
Now, I’ve actually experienced this with 3D make-up,(foam), and to a lesser degree, modeling and animating. Ruramuq brought up Beowulf. The error in the directors process as I see it was in the stage-play type movements. Stage actors need to express themselves with grand over the top movements. Film is more about subtlety of movement. Mocap is great, but take a look at what the animators did with Gollums face. Subtle. Human-like. One of an actors greatest accomplishments is to do nothing on film.Even then there is a performance. As far as the “Uncanny Valley” is concerned, I’d say we are almost there. Now that Zemekis and co. have overcome the dead eye problem, he should, in my opinion, let the animators do the faces. Cripe, Shrek has better lip-sync than Beowulf.
E.