CGI

Nova has run several recent programs about prehistoric life using CGI to create life-like dinosaurs and other creatures. Hope this link works:

www.pbs.org/video/dinosaur-apocalypse-the-new-evidence-rchsjr/

Probably cost too much to do it here, but imagine a CGI superheroine with bullets bouncing off her naked breasts, flying, doing super feats....

--Brantley

AuGoose wrote:

There are studios good enough to do it, but yes, it's pricy. Particularly because our brains are really good at recognizing other humans, so most animation without good mocap ends up looking terrible. The new Halo series is big-budget, and it still has moments when the Armored Spartan CGI is just cringeworthy.

Are development efforts looking promising to eliminate the "uncanny valley"?  My presumption is that its just a matter of further development and sophistication of graphics software to get there, but is the issue more fundamental? Such as not having a theoretical method for animating realistic humans without motion capture? 

In other words, is this a problem of knowing what we want to do, but not having the computer horsepower to efficiently do it?  Or is it that we don't know how to do it regardless of computational resources?

Shadar

AuGoose wrote:

One of the root issues is that the three-D models are rigged on skeletons. We can sculpt and position a figure with ludicrous precission. Which is why still renders can be so amazing. what the rigging doesn't have is musculature - so when one body part tenses, there are no corresponding movements THROUGHOUT the body as the whole system works together. Reality is when you extend your arm straight out to the side, there are reflexive muscular motions all the way down to your toes to maintain balance. For that, there is no (widely prevalent/commercially viable) model of the human body. And thus hand positioning or machine-learned positioning of the bones and bones alone ends up looking like very pretty stick figures without mocap - because it's actually stick figures that are being manipulated.

Very interesting. And given our eye is fantastically well tuned to the motion and body language of humans, it sounds like an entirely different model for doing CGI motion needs to be invented. 

But dinosaurs are easy, because humans and dinosaurs never even came close to co-existing, so we have no "wiring" to judge how they looked, and no genetic knowledge about how to read them for survival purposes.  Yet we can unconsciously learn a lot about a person from a few glances, or even a single one. That has worked its way into our DNA given that is an evolutionary skill. Those who lacked it didn't reproduce as successfully.

Shadar

Mo-cap eliminates a lot of the animation problems (even though some of that is still tuned using an animator).   Like AuGoose mentioned, unless your 3D model is amazingly crafted (in that it has proper bones, muscle, tissue, skin, etc. layers), then it will still look slightly off if you look real hard for it. 

Hard surface models are perfect now. Buildings, cars, planes, etc. all are as good as real life.  You can't tell the difference, and you don't even need to be using state of the art machines or software to achieve that.

In terms of "uncanny valley", what seems to be emerging in vfx is a combination of 3D modelling and deep fake technology combining.  Probably one of the best examples of this was in Mandolorian season 2.  They used that combination to recreate young Mark Hamill as Luke.  It looked pretty damn good (much better then their season 1 effort in the final episode).

Having said that, the vfx team still had to use "tricks" to achieve that effect. For example, the choice of how the shots were framed and lit were considered before any rendering took place.  Its a lot of work and still has limitations, but soon (very soon imo) the tech will exist where a digital actor stands next to a real actor and you cant tell the difference.

Digital doubles are used ALL the time in actions scenes and we don't notice.  Nearly all car chase sequence are 3D renders now (even on mid and lower budget movies).

We're getting there.

Peace.

/K

kikass2014 wrote:

Having said that, the vfx team still had to use "tricks" to achieve that effect. For example, the choice of how the shots were framed and lit were considered before any rendering took place.  Its a lot of work and still has limitations, but soon (very soon imo) the tech will exist where a digital actor stands next to a real actor and you cant tell the difference.
 

Very soon, a digital politician can stand next to a real politician, and nobody can tell the difference...

kikass2014 wrote:

Mo-cap eliminates a lot of the animation problems (even though some of that is still tuned using an animator).   Like AuGoose mentioned, unless your 3D model is amazingly crafted (in that it has proper bones, muscle, tissue, skin, etc. layers), then it will still look slightly off if you look real hard for it. 

Hard surface models are perfect now. Buildings, cars, planes, etc. all are as good as real life.  You can't tell the difference, and you don't even need to be using state of the art machines or software to achieve that.

In terms of "uncanny valley", what seems to be emerging in vfx is a combination of 3D modelling and deep fake technology combining.  Probably one of the best examples of this was in Mandolorian season 2.  They used that combination to recreate young Mark Hamill as Luke.  It looked pretty damn good (much better then their season 1 effort in the final episode).

Having said that, the vfx team still had to use "tricks" to achieve that effect. For example, the choice of how the shots were framed and lit were considered before any rendering took place.  Its a lot of work and still has limitations, but soon (very soon imo) the tech will exist where a digital actor stands next to a real actor and you cant tell the difference.

Digital doubles are used ALL the time in actions scenes and we don't notice.  Nearly all car chase sequence are 3D renders now (even on mid and lower budget movies).

We're getting there.

Peace.

/K

I hope you are right. I would really like to live long enough to see digital actors that were indistinguishable from living actors, because once you get there, you have no limits what a digital actor could do while remaining completely believable. 

Somewhere down the road such software will surely be cheap enough for hobbyist use. Imagine all the artists here on SWM if their 3D digital full-motion AI-enhanced CGI was completely and utterly realistic? If they could bring a fantastic character fully to life, at least as much as can be represented on a screen. Think how many more people could be creative if the AI CGI was so powerful it was easy to use. The content produced would be amazing.

At that point, a meta world might be a fine place to live. The simulation expands to encompass your own life. 

And then, a fair bit further down the road, all of that gets downloaded to an android. You get to meet your character, and it all seems real. 

The future fans of this genre are going to have a REALLY good time. 

Shadar 

 And then, a fair bit further down the road, all of that gets downloaded to an android. You get to meet your character, and it all seems real. 

The future fans of this genre are going to have a REALLY good time. 

Shadar 

For fans of the late 70's & early 80's shows, it would seem certain to say that the CGI Android Download factories creating Lynda's Diana Prince & Wonder Woman characters (buy one, get one free) will be working with a great amount of overtime shifts.

Monty wrote:

 And then, a fair bit further down the road, all of that gets downloaded to an android. You get to meet your character, and it all seems real. 

The future fans of this genre are going to have a REALLY good time. 

Shadar 

For fans of the late 70's & early 80's shows, it would seem certain to say that the CGI Android Download factories creating Lynda's Diana Prince & Wonder Woman characters (buy one, get one free) will be working with a great amount of overtime shifts.

The key is "indistinguishable from reality". Of course, what does that mean when most of our characters are fantasy?  

But we know EXACTLY how a person should move, how their body language should work, and we can imagine their vivacious personality based on real people of the same age and gender. We know how muscles move as they work. Getting that much perfect without motion capture will be an incredible undertaking. 

Once that is solid, then adding in flight, super strength and so forth will not trigger the uncanny valley since we are adding new elements to the experience of humanity. I think that will offer more freedom and possibly different legitimate ways of depicting it.

But to Woodclaw's point, getting weight, inertia and so forth right is going to take work too. But we can easily imagine that once we move beyond human limitations.

But then there is the new stuff. The superhuman stuff. The very pronounced muscular exertions of a young woman with extreme strength, dressed in a tiny costume, as she lifts the front of a massive Army tank, her feet cracking the concrete roadway as they prove too small to support fifty tons or more. And then a slight lifting of her skirt, cape and long hair as she begins to fly,  perhaps a faint shimmer about her as the massive weight of the tank slowly lifts higher as she begins to gather speed until she finally throws it forward to tumble fifty meters through the air.

It's massively heavy, so when it lands the ground tremor is so extreme it knocks people off their feet for a hundred meters in every direction, and the ring of steel on concrete and the crash of collapsing machinery fills the air. Then a blast of arc welder-bright eyes dazzles everyone as the air crackles when the beams strike the tank, the steel giving off tinging and pinking sounds as heats rapidly through red then orange and finally white hot until the ammo inside cooks off in a tremendous explosion that blasts her hair and cape backward, the ball of flame encompassing her for a brief moment before.

Leaping from the ground, she catches the huge turret as it falls, and then turns to walk back toward the observers, holding it high over her head, her feet crunching into the concrete with every step, blonde hair flying about her, silky strands half covering her face as she can be seen smiling while enjoying the sensation of using her abilities before such an appreciative audience. 

Yeah... I'm ready for that kind of simulated reality, especially when it gets to androids that are convincingly a flesh and blood person. 

Surely I won't live to see the last, but I'll be happy enough to see the rest of it on a screen or in some holographic kind of device or VR to simulate 3D. 
Shadar