I'd like to hear opinions on what I somehow noticed maybe not in every but many recent games as a trend. I mean not an exactly same "uncanny valley " people discussed decade ago about Beowoolf and Polar express and imo related mostly to animation but rather more subtle, more general thing a picture get after crossing certain resolution and texture resolution threshold we I believe recently crossed .
Before , around 2015 -20 , picture in games pretended to look realistic still retain some clever constructed appearance , an artistic selection where and what details we should see , a certain focus to silhouettes and mid frequency details , trying to hide what better should be hidden . A 3d grass for example kept certain level of abstract unspecific perception, the tree looked more like big mass of foliage. In a word its perception was still rather bit stylized even in games pretended to be photo-real. Like how you see it through slightly squinted eyes, seeing things more as a whole rather than every leaf. It helped to hide all the drawbacks of rasterizing render , many things online render is incapable to do .
Now I see a sort of honest and extremely procedural approach looking sort of very neat at the same time. Everything is down almost to microbes while lacking that clever touch , the selection , everything procedural with curvature masks based edges everywhere . Every grass blade, every leaf while the grass blades still casting nothing to each other , no surface micro self-shadowing which was OK in lesser resolution but now became increasingly eye catchy . Same scene looks perfect in path tracing render probably but typical on-line rasterized render makes it look sort of dry and "uncanny" art style wise. I long tried to find what it resembles to me till randomly find a well known "rainbow" portrait of Elizabeth 1. Full of tiny details done with perfectly same attention in a certain mastery manner but yet having that uncanny dead feel.
Replies
Our tools are more advanced than ever before, but tools are still tools, and require the skill to use them effectively. This is partially why Unreal Engine has a reputation now for putting out unoptimized games. Lower skill floor means practically anyone can make a game, but doesn't mean anyone can make a *good* game. Mix that with an already fairly high base performance overhead, AI upscaling, nanite, lumen, and a bunch of other features you have to modify or disable...
- Lack of technical capacity of engine (or performance and time constraints when authoring)
Engines are very good at displaying an staggering amount of tris now, have better lighting solutions, high resolution textures and more. Yet as pointed out the blades of grass still don't affect each other. Objects exist mainly in a void, they don't deform upon contact, feeling hollow and weightless. Reflection and transmission and light bounces are still fairly lacking, it's like every element exists in its own context. We notice that lack of a shared atmosphere, the absence of weight, the hollowness of the meshes when they bump against each other as if they were made of overfilled balloons.
Technology marches on and I expect implementing those missing interactions should become cheaper in the future.
- Prioritization of texture details at the expense of everything else
Likely caused by either lack of time or knowledge. Every character has incredibly crisp pores but not all have a skull under facial fat and muscles. You can't tell there are areas with shallower tissue because everything is oddly smoothed out, without a hint of the harder structure getting closer to the surface under the soft tissue. Necks might still be pretty cylindrical, no trapezius in sight. And it goes on. A lot of attention goes into texturing medium and micro details, and props got better when people started to pay attention to edges and wear patterns so objects aren't all smoke-bombed and symmetric, but there's still a lack of care in transitioning from harder to softer regions in the same material, and other types of imperfections inherent from manufacturing (eg seams on plastic, variable thickness in inset details). It's time consuming work, and it requires the artist to know how to see the world when looking at it.
This gets compounded by loose control on hierarchy of details. There's a point a surface stops being detailed and becomes just busy-looking. If everything in the scene is as busy it feels simultaneously overwhelming and still not enough. IRL we can't absorb all details all at once at top resolution. Even photos and videos can't capture everything. So when you have a 3D at super crispy busy detailing levels, it feels off. Holding off on detailing, suggesting or omitting certain ones so there's a better flow so what's detailed both stands out and still feels part of the whole is a conscious artistic decision that isn't always taken when people are stuck in the "more details create superior results" mindset.