I watched Rogue One today and I was wondering if the “his eyes look fake” assumption that people usually make when seeing a CGI character holds any water. My thinking is that the skin shader is far more difficult to nail than the eyes and that maybe the train of thought is “this looks in a creepy way fake, it must be the eyes” when in fact it’s the skin, or something else.
What do you think?
It was the skin, look at the actual Tarken side by side to the CG one. The CG model looks too waxy and is missing all the necessary subtle detail and pigments. Animation of the lips was also a little off. They also exaggerated the facial features/structure too much.
I’m looking forward to seeing CGI Ben Kenobi in episode 8 though, since he’s a force ghost and as such is blue and semi transparent, the skin problem is lessened.
Lol, blue and semi-transparent…
I can’t find a good image of CGI Tarkin on the net yet and my contact-lenses wee acting up in the movie theater so I couldn’t get a really good look. Leia however due to the better lighting was obviously fake. I mostly found the lips to be spoiling the illusion.
I did notice General Tarkin’s waxy skin in some shots as well, did Disney not use their own Burley SSS algorithm for the skin material (I would’ve thought they would make use of their own research in these movies)?
All of the hard-surface CGI though looked good, though it doesn’t have near as many things to take into account as a full CG human.
Generally in holywood CGI, I notice they tend to make eyes far more white than it should be. The “white” part of an eyeball is usually not white but a very very light greyish tone and as people age their eye “whites” start to lose their color an turn to more reddish/greyish tint.(Depending on the racial background or living conditions it may be yellowish/brownish and darker than usual). In the film there was too much contrast for me to differentiate this natural look.So it is passable I guess (since I doubt I can do a better job).
About the skin it was a little bit to glossy at some angles.
One thing to note though is that in the original star wars the lighting setups, lighting equipment and the bath process (and chemicals) for the 35mm film are very different than todays digital processes. All this different processes give 60’s -70’s films a certain look and feel. This makes a huge difference in my opinion.
There needs to be a Turing test for cg characters someone needs to pull a Simone and see if people notice(I think this happens but it needs to happen on a bigger scale). I think when a character is obviously cg then people react, it’s terrible, it’s horrible all the canned response come out. We know Peter Cushing was dead at the time and the late Carrier Fisher was 59/60 at the time for this movie been made so obviously there are CGI characters.
The reason I personally pay no attention when people pontificate over CGI humans is that most people can’t spot the digital doubles that Hollywood doesn’t do a song and dance number about. I think most people have seen more digital doubles then the actually know it’s only the obvious ones that seem to get everyone’s panties in a twist.
Digital doubles are one thing (they are usually seen at a distance or during kinetic action scenes) but digitally recreating a human performer we all recognize is quite another. Especially when it falls right in the middle of the Uncanny Valley (which I feel both Tarkin and Leia did). They’ve made huge strides over the years, but it still falls short. And it was the eyes. They looked glassy and dead. Avatar managed to make digital humanoids with realistic facial movement and eyes that felt alive, so I really wonder why seven years later we find a top shelf company like ILM still struggling with it.
Maybe just because they were humanoids we are willing to accept “slightly off” as good enough, i.e. that uncanny valley is really narrower when not-really-human creatures are involved?
Watching a YouTube video that explained how Tarkin was done (finding a similar actor and mo-capping him) I really wonder if that had a proportional benefit for the cost versus say, investing in creating a better SSS shader or even IDK, training a neural network to pull pixels off old footage with similar lighting and transforming them.
The thing that struck me about the R1 characters was that their proportions just seemed a little off.
Tarkin’s face seemed to long and narrow and he features just a bit too exaggerated (almost like a caricature). Leia’s eyes seemed just a bit too far apart.
Given how well we know these characters - this is what stood out for me.
one thing to mention:
if you know its not real you looking for something to proof it cant be real!
i believe that most people who say the movie didnt notice that those 2 characters where full CG.
yes, i had some shots of tarkin where i thought “something is off” but on other i totally forgott about it.
and i dont believe that one of the best vfx company in the world, the only one who knows every trick from star wars, one of the best vfx supervisors didnt check against all kinds of reference material to give you the best possible result.
Fully CG humans in a film environment is known as one of the hardest effects to pull off, as there are a large number of things you need to take into account to avoid what is known as the uncanny valley. To this day, it’s something that only a handful of productions managed to achieve full realism in.
Of course, our perspective on things could also be skewed with the heavy dose of color-grading applied to the average film (usually bringing an orange and teal veneer). It’s to the point where movies aren’t necessarily the best references to go to in terms of photoreal materials and lighting.
i know its the hardes thing to create full screen digital humans and i’m aware of the uncanny valley.
what we see in “Rouge One” is probably the best result ILM could create in the restrictions of time and money.
surely they could have improved everything but then the movie would never hit the cinemas.
Hm the eyes did bother me less than the skin look. I was quite surprised about that they were not able to get those done better.
The animation seemed fine to me. Some mentioned that the gestures was artificial.
They must have been passing around the pipe at Disney when they made that decision. CG can’t replace human actors, the slightest blip of artificiality will be noticed and break the whole suspension of disbelief which movies rely on. Tarkin should have been a hologram at best.
The thing that always bothers me the most is always the iips. They´ve just not found a way to do believable lip motion yet… as for Grand Moff Tarkin, I haven´t seen any “live motion” example yet as I haven´t seen the movie, but even stills look CGI to me, which I wouldn´t have expected from ILM… I think we´re at least 5-6 years away from achieving realistic CGI humans, but then again, I probably would have given the same number estimate 5-6 years ago…
Maybe there’s room for a “Tarkin Clone War” challenge for the Blender artists here… Probably need separate still and movie categories given the complexities.
Finally got around to seeing the movie. You know, technical aspects and uncanny valley aside, I wonder if the mocap actors are partially to blame for the animation feeling artificial. Knowing that the movement of the little dots on their face are what drives the animation, they may have a subconscious tendency to over express, which then needs to be manually adjusted. Just a thought.
I don’t think it’s a lack of capability with all the resources big studios have. (The area of the uncanny valley may be getting narrower.) Rather I think it’s working down to pipeline and budget constraints. To get things to happen within the production window they’re given for a movie, shortcuts happen in some places. (More details, more fidgety work, etc. Stuff like that could add another year which isn’t going to be covered.) So something like a movie that specifically showcases CGI characters from 5 years ago might be better at doing it than the CGI characters in a current movie where they’re occasional stand-ins or cameos. It’s just a matter of where those in charge of making the movie consider where the effort should be spent.
Not a criticism of those working behind the scenes. The people good enough to get these jobs likely work as hard as they can within the limits they’re given to work with.
Good point. At some point the Director just has to say “good enough”. Just look at the Digital Emily Project. That was 6 or 7 years ago, and I still think it’s the best example of animated CG human to date. The difference is that team didn’t have the constraints of a production deadline.