@@Harvester88I don’t understand how Rockstar made some of the most fluid looking character animation in Max Payne 3 and then they (nor anybody else) ever tried to expand on it. Max Payne 3 animation-wise is still imo the best feeling third person shooter and it’s in large part due to how well animated Max was.
Exactly, I get turned off by these big budget crazy looking games that the second you want to shoot any object in the scene it doesn't move an inch. So clearly breaks the immersion, me being a prop that moves through a pre-built circus and can't interact with it in any way than at predetermined points.
Trust me, engineers want better realtime physics. Complex, interactable, and accurate physics systems are just an incredibly difficult problem to solve.
Agreed, I want a Battlefield type game where the destruction actually feels real and not like “oh, I clipped this wall so now the whole thing is turns to dust instantly”
Let’s put it this way: Cinematography in films and shows have always been path-traced. They have continued to get better and better at how to light and shoot cool looking scenes using those tools. Games will just have to start employing the tricks of the DP of a film for how to make stuff look better and better using path-tracing. Hiding extra lights and enhancing practical lights with hidden ones, etc.
@@RicochetForce Snow in the Desert was that same studios follow up and is absolutely mind blowing. It’s the first time in my life that I have been completely fooled where I thought they had full motion video actors mixed with the CG ones in some of the scenes.
@@itsd0nk Exactly, d0nk. It took us decades for realtime graphics to start matching (or surpassing) offline renders from the PS3/360 era. Our curren games are decades from matching something like Snow in the Desert in realtime. And that's just on the visual end of things.
Yep. One of the cool things that will come as the tech matures is seeing devs and artists start to fully utilize simulated lighting techniques as fundamental part of the creative process, rather than just as a supplement to more traditional techniques.
Everyone focuses on visuals, but what I truly miss nowadays is real interactivity - physics-based props, environmental reactions to elements like fire, advanced water simulations, improved AI for NPCs, and such. Far Cry 2 was a huge leap in these aspects back then. Everything that followed? Just boring. Wish more people would talk about that.
Games back in the late 90’s and early to mid 2000’s where doing that but the Industry dropped that to focus on linear shooters and repetitive open worlds.
Yeah, improvements to the simulation of a game are what excite me the most. You can see the cracks when you have static ai, a world that can't be interacted with, loading screens, and repeated animations, among others. It is a difficult thing to do, so advancements towards making it easier would be great.
YES! AI! That's tech as well, and a big part. We got to a point that tech analysis of games is like "well the texture filtering could use work". Brother, AI, physics, animation, sound. So many spots to improve in.
Real-time character movement without having to motion capture. Real-time facial expressions. Hair. Skin. Background character fidelity. Actual frame rates and consistent frame times. Real time cloth sim.... There are so many things that aren't even remotely close to any ceiling. The lack of points to critique was within the confines of todays standards. Doesn't mean there isn't room to improve. We are decades away from actually achieving fotorealism no matter what the marketing people are spewing off.
There are already plenty of games that don't use motion capture to do their animations. This tends to apply to non realistic styled games. Plus things like animation blending are a thing
"Real-time character movement without having to motion capture" "Real-time facial expressions". You do know that the reason we capture motion is because that's the only way it will look realistic? Before motion capture everything looked cartoony because it was done by hand. You are probably going to say "i don't want them done by hand, i want a character that reacts realistically to this audio file and this text file"... once again, this was done already and it still is outside of cut scenes. The face rig will react to the voice of the actor and the results are pretty meh. Capturing a real person is the only way of delivering what the developer intends for you to see. Even if an AI advances to a state where it can react automatically to a pre-recorded sound file and do a good job, it may never do what the dev intends because there are lots of ways of reacting to the same line. Unless game development in the future will be done with AI actors who literally do takes, and the dev says "CUT, do that with more emotion.. scene 3 take 5, ACTION".. why not just ask a real actor? inb4 an AI doesn't need to be payed.
@@C--A That's because you can only capture so much data. Plus there are plenty of other things you have to simulate besides the animation on the surface. Plus you never use raw motion capture data as it needs to be tweaked by hand. In combination with that data you also have to simulate the movement of muscles underneath the skin to make this stuff right. If you've ever seen a model used for animals in movies you'll see that the model isn't just the outside but a full muscle and bone simulation underneath the skin as well
My take on this is that the biggest benefit of path/ray tracing is in giving indie devs and small studios the ability to produce great graphics (mostly lighting) cheaply. Rasterization-based techniques work very well, but it takes a lot of time and effort and expertise to employ them effectively. With path/ray tracing, devs just put their objects and materials and lightsources in the world and the algorithms do their magic. We'll see improvement at the top end, sure, but the big impact will be in raising the bar for the low end.
Yeah , i'm not a game dev but i have used tools like Unreal engine and Nvidia omniverse ( mostly isaac sym for robot simulation ) and they're a god send for how easy it is to have good looking ray traced lighting . Ofc optimization is still required and needs a lot of tuning but that isn't a big issue since u have to do that in any case
Hoping audio path tracing and better audio-materials interactions become more prominent. Also hoping fluid sim/animations (primarily wind & water) see big refinements. The fluids animations in Avatar stood out as being extremely simplistic relative to the game's geometry and lighting fidelity. Physics/animation systems to improve model interactions and reduce/eliminate clipping/sliding would be great 3:27 Definitely an important to point to remember as well.
And rendering subdivision surfaces (NURBs, Bezier, Catmull-Clark, ...) and displacement mapping directly in order to improve LoD, memory usage and spatial subdivision would be awesome, either via adaptive tessellation or raycasting. This is somewhat related to simulations since those can also benefit from the use of objects not based on triangular meshes.
Show me a game that REALISTICALLY portrays a character actually DRINKING from a CLEAR MUG! And we're pushing 4K res/textures running at 120fps, but nearly every game still seems to suffer from CLIPPING/COLLISION issues. Also, ANIMATION still seems to have a LONG way to go until characters consistently appear to be properly interacting with all elements of their 'environments'...
@@HiN00bs12 Does it actually have consistently 'perfect' collision/clipping and interaction with ALL the foliage in the environments? I know at least some of the plants appear to react to characters moving through ground vegetation, but most games seem to be inconsistent with it at best. Guess I'll have to take a closer look at footage of Tsushima! I still remember being amazed when the FIRST Assassins' Creed footage showed him actually using his hands to push through crowds, but haven't noticed things progress very far past that in the last decade...
@@StreetPreacherr ye, in the first link below, others are also instances of really cool food animations. ua-cam.com/video/t5csS7xcEzc/v-deo.html&ab_channel=ThePeteriarchy ua-cam.com/video/_dOPdLmucJg/v-deo.html&ab_channel=BrendonCasler ua-cam.com/video/MHYs18YRDdY/v-deo.html&ab_channel=dacyl
Absolutely! The first thing that comes to mind is the ubiquitous use of advanced physics simulations. Remember The Witcher 3's Hairworks on Geralt and monsters? Or Batman Arkham Knight's interactable smoke, paper, and rain from Nvidia Gameworks? That stuff seems to have fallen to the wayside, probably due to being computationally expensive, but it looked awesome. Both TW3 and AK runs great now on modern GPUs with everything maxed out at 4k60+ on a 30/40 series card. Imagine a game where literally every visible item of cloth is actually a cloth sim that deforms accurately according to the geometry touching it, and all the traditional sprite based particle effects were replaced with a Gameworks style physics simulation. I think that'd go a long way to ramping up immersion.
I believe that path tracing in its current incarnation has plenty of room to grow. We don't even have lighting caustics yet where light refracts through transparent glass and water, causing it to scatter and collect in ways based on the medium of the geometry and the geometry's makeup. But, as soon as you have every lighting phenomenon that the human eye can observe be systematic and real time, the only limitation to the realism a game could have would be the model and material assets itself, and how they behave.
I would also like to bring up, that the question itself in the videos title is kind of loaded because it makes the implicit assumption that realistic graphics means better graphics. HiFi Rush is not a game that I would consider looks realistic, but I think it has great graphics because it has visual technologies that allow itself to look like it was an anime or a cartoon Edit: just got to the part where john mentions this! Nice
The one thing imo that could truly make today's games feel outdated in the future is breakthrough advancements in NPC AI. While it's not really a graphics factor, it is certainly an aspect of what you see happening on screen. Personally I've always found NPC's just standing around like lifeless mannequins repeating the same lines over and over very immersion breaking.
A bit surprised few people seem to mention artistic input. Sure, there are the purely technical aspects, but the general community doesn’t seem to understand where the artist and developer input make the difference. Have a look at the varying results offline rendering for architecture has brought us for at least 15 years. Archviz depended on path tracing almost exclusively for a long time
Maybe it is just me but I believe more progress should be made in animation. It can't be that we are approaching photorealistic graphics and yet characters in videogames still move in a very unnatural, unrealistic way
RoboCop is AA game I am Polish so I was watching Interview with Teyon developer, he said animations of characters looks like that because of the costs. Smaller company’s doesn’t have such huge budgets.
Graphics have a long way to go. Even if path tracing solved all lighting issues(which it doesn't,but its close) we still have to solve all the others ,like texture quality, assets quality, physics and animations. This is the future, I think with the adoption of 4K ,textures have improved vastly and will continue to improve in the future. Physics however are one place where even Alan Wake fails miserably, this is going to take several decades to perfect.
Nope. Only till the end of this generation. The only thing that will need to improve are CPU’s with physics. Graphics technology have already peaked, we just need the budgets to show us good looking graphics.
And the graphics tech I’m talking about is virtualized geometry. It basically means graphics can be infinite in detail. It’s already been fully implemented in UE5 and we just need to wait for other companies to implement the same tech in their own engines.
It these goobers in the video are to stubborn to acknowledge the potential of this tech because it completely contradicts like %100 of what they say in every god damn video they’ve made in the past few years lmao. And it also means that consoles have the possibility to literally outperform PCs.
Honestly, I am okay with how path tracing looks like now but what takes the immersion away is smoke, water and cloth physics I really hope the next push is for real time fluid and cloth simulation
Water physics won't have a solution for a long time. Fluid physics are an especially complex area to compute and to get right without eating up all the resources you have
Path tracing is the least of it, graphically. Sure, lighting is amazing, but what about everything else in the visual pipeline. We still have a way to go.
Art direction is big. I was looking at the BioShock infinite DLCs, you can tell it's old game, but it still looks great. It's because of the art direction.
Also worth looking at PBR / rendering discussions that happen at Siggraph. Even some of the multi layer translucent scattering work that was being chatted about 10 years ago at that conference will be doable in tech like Substrate in Unreal in the future, but right now an extra material layer that interacts properly can add about 4-8ms on a 2080ti for a full screen asset. But at any rate, current day siggraph film talks are always an interesting look forward for games :) Outside of more complex materials: We still have harsh limits on number of shadow casting lights. Transparency is still expensive and not fully featured in most engines. Simulation for visual effects (volumetrics sure, but even water sims, remeshing, etc) is still very expensive. We use lights and textures that have 3 channels (RGB) instead of a channel per wavelength, or some form of spectral rendering (like in the Manuka renderer). We're probably a decade off having software we could even use to make that content. No doubt there will be plenty of big interesting leaps like path tracing, but there are 1000s of small things we work around at the moment that will benefit from better hardware :)
@@vitordelima Hmmm GDC maybe? (game developers conference?) Siggraph does have a lot of real-time work these days too! The "Real-time live" sessions are great, often have talks from big game studios, really interesting tech stuff, etc :)
Harsh limits on the number o shadow casting lights is an issue solely caused by rasterization. The more lights you need to compute to calculate shadows the heavier it gets. Ray tracing and path tracing do not have this light source limit
The more realistic games start to look, the more boring they seem. I want games looking surreal and/or 'like games'. I agree that as graphics become more complex, it is more and more difficult to make everything coherent and fit together. When you raise the quality of one asset, you must raise the quality of everything. Making retro-DOS-looking-games is fair easier to do in that way. Those games are much simpler in graphics and gameplay hence the reason you also don't really see any bugs. I think back in the day when 3d games became a thing, the graphical quality went way downhill. Old 3d games didn't age well at all, whereas old 2d games did.
Art design is what makes games look good. Path tracing is just like any other tool that needs to be used properly to generate an aesthetically pleasing look. Better tools don't always mean better graphics. Case in point, modern Hollywood has way better tools for CGI, but most movies from 2010 until the present look so incredibly fake compared to movies from 1990 - 2009. The original Jurassic Park had primitive tools for CGI, and it still looks more realistic than any of the Jurassic World movies.
I think one point where computing in general still has a lot of room for improvement is real-time fluid/particle simulation. Things like waterfalls, clouds, and destruction in general still feels very "that's a video-game", Even with other visual aspects becoming more realistic every day
You can find the answer in 3D animated movies and see what has been accomplished over the last (almost) 30 years, rather than AI generated games. The main, and most important, thing remains: the artistic style. PS AI generated games aren't new, I remember Sentdex having shown what how an AI represented GTA 5. But I'm not sure that what he showed was realtime or not. And that has been YEARS ago.
Automation of processes, modular plug and play of assets and using generative A.I. to apply itself to visuals will move things forwards. But at the end of the day, it's the story its the playable mechanics that are going to define games. It's not the case for visuals any more to determine how good a game is.
The next algorithm to get working in realtime is Spectral Pathtracing. But having enough rays per pixel to drop “denoising” is gonna be the biggest hurdle! Thats when realtime converges with a Weta shot (strictly wrt lighting - accurate af muscle/fasciae + cloth simulation is a whole other thing - Also we have yet to see UE5.3 VDB implemented in an existing AAA title)
Maybe it’s just me but when given the option to disable ray tracing, I do. Not for the performance benefits, I just don’t like how it looks. And this is true for every game that offers it. I just feel the image is more satisfying to look at without it. Brighter and more colorful. There’s another part of me that says I can go outside and witness real time ray tracing. I enter a video game world to escape reality. I also feel this emphasis on implementing RT is forcing devs into utilizing “hot tech” rather than focusing on interesting game worlds. It makes me cringe when I hear the phrase “but it lacks any sort of ray tracing”.
In order for things to really advance, developers need to cut out some workloads. So shift entirely into Ray Tracing, leave baked lighting in the past that way they are not doing 2x the work to light scenes for two different types of lighting...its just a waste. Same with Mesh Shaders, leave Vertex in the dust and move on. Gamers WILL buy new hardware if forced to. But we can't move on if we are still doing "hybrid" development for a huge array of hardware, which I can only imagine adds to the cost too. Not everyone gets to have cake and eat it at the same time as everyone else.
graphics look like real-life but animations still looks like 20 years ago, we need more active ragdoll animations (euphoria engine) in games, like gta 4, 5 or red dead
Yeah it's hard to imagine where Graphics tech goes from here but logically Ray Tracing is the mathematical accuracy of the behavior of light the same could be done for all aspects of the physical realm physics material and liquid fog smoke and fire could all get the equivalent accuracy
I think at a point in the near future (5-7 years, probably before any kind of major graphical improvement beyond path tracing), games won't have traditional graphic engines and the constraints of hardware or programming. I think A.I. will be producing the graphics in real time in games and movies. A.I. artwork and animation is improving literally weekly now. Things are getting creepy; just from the perspective of how fast certain aspects of A.I. are accelerating.
Realistic physics and destructible environments. There was a kind of niche period when people were trying things with PhysX on GPUs in the late 2000s, but it didn't catch up en masse. So after path tracing becomes kind of norm in the late 2020s, just as dynamic shadows did after the early 2000, and SSAO after the late 2000s, it would probably be physics. Maybe they'll add a separate PPU (physics processing unit) onto the chip, or just let tensor cores procedurally generate damages/interactions. Because textures are basically already up there, maybe need just a tiny bit of polish when 24~32+GB cards become the norm instead of 8~12GB. Also, draw distance. It's difficult to notice in motion, as you just play immersively, but when you slow downd a notch and start to look for it, it's easy to notice in every game with large open locations, like Cyberpunk for example. Actual supersampling. No matter how much you tweak temporal algorithms / DLSS / XeSS / FSR, if I have the performance of a hypothetical RX 14880XTX to 4x4 SSAA on top of a 3840x2160, I will choose it any day.
Textures, assets, animations, physics, etc. Can definitely improve. So, even if we perfected lighting and reflections, we still need to work on that stuff.
On the graphics side I want better animations, real hair strands, realistic expressions, more geometry, and honestly more lights. By more lights i don't mean adding them for no reason, but light shafts that go through cracks in the wall. I'd also love more physics based effects like smoke, fire, water, etc. Light interacting with this would be great.
My benchmark for graphics is "what would someone who doesn't play videogames instantly pick out?". Sure, we've got traced rays coming out the wazoo and high poly characters models with sub-surface scattering. But at the same time, we've still got issues with accessories clipping into characters and the ground is just a flat texture with some 2D cut-outs of grass scattered around in tufts.
Currently a lot of image is still rasterized and old tricks are still used. Most of the volumetrics and transparent surfaces will need to get RT to fix a lot of the problems. Would love to see water simulation with proper refraction and IoR for next Subnautica etc. Simulating the camera and lense system for shooting rays from lense and sending rays at different points of time would solve depth of field and motion blur.
Honestly as a 3D artist we'd love it if we no longer needed to optimize a low poly mesh. It would hasten the workflow quite considerably. I'd imagine as CPU and GPU get faster we won't need normal maps anymore for those smaller details.
We'll eventually figure all those things out. I'm more interested in if we'll eventually move on from polygons as the main medium, just like we moved from 2D to 3D 30 years ago. Something like micro voxel rendering at a high enough level would be really interesting to see, especially for physics simulations. Or are polygons along with all the smoke and mirrors we invented just the defacto most optimal way to do things by now.
I think next improvment must be in physics department. Developers just improve image and they forget physics , nut they for me is more important. Zelda TOTK prove that.
I think realistic visuals are just a stepping point along the way, stylized or hyper-real rendering will be the next step. Games where the visuals are driven by the style of the game and not by real life always have a leg up already imo, so just aiming for photorealism isn't an endpoint.
Path tracing and Photogrammetry will because close to standard by the end of this decade. Physics and animation need a HUGE upgrade and I think VR will also be next in terms of gaming but at least 15-20 years before that’s standard.
We still have much distance to cover from being able to do current real time ray tracing in 60+ fps to achieving quality of what would currently take render farms days per frame to produce.
Once virtualized geometry and full path tracing is common, further improvements will still occur, but probably with diminishing returns. Real time graphics will then improve similar to how pre-rendered CGI has improved in the past years: At a much slower pace. Then the main thing that will influence the graphical result will be the art style and the quality of the game assets. The latter will largely depend on how many artists are working on them. Just like in CGI movies today.
Nanite has recently been fully implemented in UE5. And it’s only a matter of time before other companies do the same with their own game engines. And with virtualized geometry bringing the capability of infinite details, it entirely depends on budget know for better graphics.
@@balloonb0y677 Yeah, though even UE5 supports virtualized geometry only partially for animated meshes. Other engines currently don't have it at all. They will probably get to it in the next few years.
Graphics were good enough five years ago, and they are especially good enough now. Make animation better. Make physics better. Make gameplay better. Do stuff beyond “look at these shiny objects that are in a world of bugs and glitches”. Make gaming great again.
The next thing after path tracing is wave tracing. Light in the real world has properties of both a wave and a particle, while standard path tracing only considers the particle aspect.
The same question was asked before path tracing was discovered. And before physics were introduced. And before shaders appeared. And before 3D took over. Etc, etc.....
To me Graphics improved when RT and Motion blur is turned off . Nobody is choosing 30 fps with RT over 60fps without RT. Shiny floors are not that important .
You really don't understand what RT does then. It's not just shiny floors that it does. It's the entire lighting pipeline that is being improved upon. So what it's expensive, other visual effects like shadow maps are still expensive to this day
I believe that once they incorporate the new AI dl models for NPC personalities, combat and the like this will overshadow ray tracing completely. Graphics are sort of at their peak and in this day and age games need something more than visuals. There are many games with breathtaking graphics but lackluster gameplay. We need something new, not 'lighting affects'. Just doesn't cut it by any means.
Yes. There is always room for improvement. As long as I can wait 10 hours for Houdini to finish a 10 second fluid particle simulation, as long as you cannot simulate the sahara with every grain of sand reacting physically correct, there will always be room for improvement. Because you can bet your butt if a triple A gamedeveloper had the resources to render every grain of sand in a desert, he would totally do it. As long as a high detailled VDB voxelcache of a nuclear blast can easily take a TB of cache space, there is things that are still years away from working in realtime. Ask those VFX guys in Hollywood about caching times and rendertimes. When it comes to complex destruction simulation with true to world detail, even seconds of animation can take up hours and days of caching time and literally Terabytes of space. Of course AI will speed up things a lot but only because you can RENDER everything, doesn´t mean you can get everything imaginable to work smoothly in realtime. Let´s just look at water. A closer look. Or think of fire. Chemical reactions like acids. Not even getting started with caustics, dispersion, diffraction, complex translucent materials. There will always be a fraction of users longing for the most immersive and realistic experience and there will always be a tech company looking for reasons to sell you an even better product.
This is a ridiculous notion and a failure of imagination. All ConcreteLlama had to do is go look at things like the featured shorts in Love, Death & Robots, or any number of celebrated animated films to show just how FAR realtime rendering has to go. CG movies have had pathtracing for ages, and yet they still continue to improve dramatically. For games, the focus should be on dynamism and interfaction. Make it so you can interact with practically every element in the environment. Build it up, bring it down, exploit natural forces, be subject to the effect of natural forces, all while robust physics engines and robust pathtracing setups support the entire effort.
Path tracing has a long way to go. Most games only calculate the surface level of lighting while complex materials for skin, cloth, metal, liquids, glass, fog etc are still a simplified approximation. There are good reasons rendering packages like 3dsMax, Maya, Blender still take more than 1h per frame without any upscale or denoising trickery.
I think when all games look super realistic, it'll be hard to differentiate what game it is just by looking at it and we will start to see more focus on the art style instead
The funny thing is that on consoles I miss realistic looking and realistic playing shooters. For me, they all have too much gimmicks going on, like the extra powers, more focus on style, skill trees, wall jumping, battle royale, etc. I wish we had more games like Brothers in Arms of Ghost Recon kinda games. But hey, I'm getting old I guess.
All games won't go for a super realistic style though. Path tracing doesn't mean you have to go to realism for your visuals. Plenty of games already focus on different styles besides realism and that isn't going to change. Obviously not every developer wants to aim for realism and that hasn't changed at all
Alan Wake 2 looks great but it would take 1 second for mostly everyone to identify that it's a game. We are very far from photorealism. Even movie studio with rendering farms often struggle to reach it.
When we look at it component wise I think our GPUS have started to reach the limit with path tracing, and I think for better realism our next target component wise is ram and cpu. We need real time simulated things now.
Physics, easily. I get kind of upset by people being so oblivious/inside the box in regards to video game visuals, means that devs will continue to upgrade what doesn't matter (textures, geometry etc.). After that, I think video game tech is officially completed. For the worlds to look photoreal and ALSO you can interact with it realistically on every level? Water, hair, cloth, destruction/deformation etc. That's the endpoint. And then again, that's an end point faaaar off. So don't worry, it's gonna take a while.
When we've hit perfect light and physics in games, and the assets/materials are indistinguishable from reality, thats when skills from movie production should transfer over, we'll see more cinematographer and set designer style roles. It'll be very freeing and much like painting, when photorealism has been achieved the next step is moving into more surreal explorations of design. interesting times ahead.
i still think work needs to be done to character models faces still they dont look real, you know its a computer image compared to the clothes there wearing were metal and leather looks almost real you seen games like unrecord that are going for super realistic enviroments hide faces because they no it will blow the realism, its just not there yet
Character faces is a combination of both lighting and texturing. You need to simulate or at least fake the way light scatters when it hits skin which is called subsurface scattering. You also need to make sure the texture work shows the necessary detail to make it look right
i noticed that when u lower the pixels like going from 1080 to 480p in gow 2 makes it look so fk reallystic even more than games like sekiro in 4k because in 4k u can see every little detail and every error they did when making the textures
Path tracing in games still not fully accurate, we need 100x more powerful hardware for things like colors from oil spill/CDs/prism, frosted glass or any material that uses light's wave property, movies resolve it by pushing millions of rays of different frequency per-pixel.
yes, tons of stuff...ultra realistic body and face for every npcs in the game world, realistic non robotic animations for everything, physics...seems gta6 will be the next big improvement in those areas according to the trailer.
Man .. I used to think half life 2 , Max Payne 3 etc had the greatest graphics ... Because the games were imeraive and good ...and our minds interpolated it into super path tracing... But I play modern day games and
Remember folks. Path tracing is a tool. It's how they use the tool that is always going to get better and better. Also Path tracing is just the beginning. Light is not a ray scattering around. Light is a wave and each color has its own wave length and is also affected by the physics like gravity as well as interaction with the infinite amount of wavelengths being scattered around like sound waves. There is so much to improve on we are just not even scratching the surface yet. But we just FINALLY! started to attempt to scratch the surface. Developers will not spend weeks and months trying to fake the lighting with raster based techniques. They will actually spend weeks and months thinking of how to better bring the lighting to their intent instead of trying to imitate what was their intent. Big change there.
CIG just demonstrated very good water physics, fire propagation, and wind effects with their Star Engine. When you see how many super immersive features they've added to that engine, the long development time makes perfect sense. Bethesda could never. 😂
I think it’s more cpu side stuff, more complex ai in the npc’s, more npc’s more simulation and physics for dynamic intrinsic gameplay etc than more shiny stuff that’ll feel next gen, rdr2 wasn’t just pretty (without ray tracing too) but had a complex world with a lot of moving parts where cyberpunk is very stiff and sterile beyond visuals
Of course, things like geometry and texture can always get better. And path traced lighting still can't recreate how reality looks, only try to approximate. Put it this way, if graphics can't recreate what you can see right now with your own two eyes, then graphics still has a long way to go.
Obviously yes. Movie VFX have been path traced for over 20 years and they are clearly (as a trend, not every example) getting wildly better year over year.
Devs need to stop obsessing with vram destorying details and focus on leveraging cutting edge technology to bring superior art design to life. Look at rage for example. Cutting edge tech brought a style of graphics way beyond its time. Fully baked in raytrace lighting. Yes static but a great first attempt
Before pathtracing the only thing getting better in graphics was the resolution. That is because the ps4 and xbox one held back the generation. Today you need at rtx 4090 to max out the newest games but in just 5 years i bet you that will be considered the base line. Whats next in computer graphics is close to photorealism but i think thats far ahead. Future games will probably be a fully functioning reality just like our reality we will be able to simulate reality inside our own simulated reality and the end goal is to escape the boundaries of our reality and enter other realitys to explore.
Was expecting some mention of realtime simulation, translucency, subsurface scattering, mocap data with procedural generated in between, morphing. Gaming is far from cg in terms of graphical details because of optimisation. A simple comparison between both will list a lot…
The graphics don't need to get better - path/ray tracing will make the current gen graphics look like real life (and more importantly, the games need to be more fun)
I still think ray tracing is overrated. You can fake lighting and still make it look realistic. I'd rather see realistic details density, realistic environment physics and reaction to movements and realistic animations especially humans.
I think after raytracing the next big thing will be proper simulations for things like water, fire, and propagation of elements like that.
animations in general too, even in new games like Alan Wake 2, the animations still looks stiff in some ways by comparison.
@@Harvester88I don’t understand how Rockstar made some of the most fluid looking character animation in Max Payne 3 and then they (nor anybody else) ever tried to expand on it. Max Payne 3 animation-wise is still imo the best feeling third person shooter and it’s in large part due to how well animated Max was.
@@jordanroose8499max Payne 3 animations don't look good by today's standards.
All that should be priority over RT/PT.
Yeah fisics will get better
Physics and deformation is what I want to see. Lighting looks amazing now, textures are amazing now, but everything is so... stationary.
Yeah, stiff, you mean?
Exactly, I get turned off by these big budget crazy looking games that the second you want to shoot any object in the scene it doesn't move an inch. So clearly breaks the immersion, me being a prop that moves through a pre-built circus and can't interact with it in any way than at predetermined points.
Ai should getting better and with destruction
Trust me, engineers want better realtime physics. Complex, interactable, and accurate physics systems are just an incredibly difficult problem to solve.
Agreed, I want a Battlefield type game where the destruction actually feels real and not like “oh, I clipped this wall so now the whole thing is turns to dust instantly”
Let’s put it this way: Cinematography in films and shows have always been path-traced. They have continued to get better and better at how to light and shoot cool looking scenes using those tools. Games will just have to start employing the tricks of the DP of a film for how to make stuff look better and better using path-tracing. Hiding extra lights and enhancing practical lights with hidden ones, etc.
Bingo. Look at something like Beyond the Aquila Rift. Obviously dated now, but it far exceeds ANY game we've seen in realtime to date.
@@RicochetForce Snow in the Desert was that same studios follow up and is absolutely mind blowing. It’s the first time in my life that I have been completely fooled where I thought they had full motion video actors mixed with the CG ones in some of the scenes.
@@itsd0nk Exactly, d0nk. It took us decades for realtime graphics to start matching (or surpassing) offline renders from the PS3/360 era. Our curren games are decades from matching something like Snow in the Desert in realtime.
And that's just on the visual end of things.
Yep. One of the cool things that will come as the tech matures is seeing devs and artists start to fully utilize simulated lighting techniques as fundamental part of the creative process, rather than just as a supplement to more traditional techniques.
Everyone focuses on visuals, but what I truly miss nowadays is real interactivity - physics-based props, environmental reactions to elements like fire, advanced water simulations, improved AI for NPCs, and such. Far Cry 2 was a huge leap in these aspects back then. Everything that followed? Just boring. Wish more people would talk about that.
Games back in the late 90’s and early to mid 2000’s where doing that but the Industry dropped that to focus on linear shooters and repetitive open worlds.
Yeah, improvements to the simulation of a game are what excite me the most. You can see the cracks when you have static ai, a world that can't be interacted with, loading screens, and repeated animations, among others. It is a difficult thing to do, so advancements towards making it easier would be great.
YES! AI! That's tech as well, and a big part. We got to a point that tech analysis of games is like "well the texture filtering could use work". Brother, AI, physics, animation, sound. So many spots to improve in.
Forgot about Far Cry 2. What an incredible game
Real-time character movement without having to motion capture.
Real-time facial expressions. Hair. Skin. Background character fidelity. Actual frame rates and consistent frame times. Real time cloth sim....
There are so many things that aren't even remotely close to any ceiling. The lack of points to critique was within the confines of todays standards. Doesn't mean there isn't room to improve. We are decades away from actually achieving fotorealism no matter what the marketing people are spewing off.
More geometric and texture detail (it's still subject to certain limits nowadays), multiple light bounces, ...
There are already plenty of games that don't use motion capture to do their animations. This tends to apply to non realistic styled games. Plus things like animation blending are a thing
@@crestofhonor2349 Even the current best motion capture for CGI films & video games. Is a long way off from being close to real life human movements!
"Real-time character movement without having to motion capture" "Real-time facial expressions". You do know that the reason we capture motion is because that's the only way it will look realistic? Before motion capture everything looked cartoony because it was done by hand. You are probably going to say "i don't want them done by hand, i want a character that reacts realistically to this audio file and this text file"... once again, this was done already and it still is outside of cut scenes. The face rig will react to the voice of the actor and the results are pretty meh. Capturing a real person is the only way of delivering what the developer intends for you to see. Even if an AI advances to a state where it can react automatically to a pre-recorded sound file and do a good job, it may never do what the dev intends because there are lots of ways of reacting to the same line. Unless game development in the future will be done with AI actors who literally do takes, and the dev says "CUT, do that with more emotion.. scene 3 take 5, ACTION".. why not just ask a real actor? inb4 an AI doesn't need to be payed.
@@C--A That's because you can only capture so much data. Plus there are plenty of other things you have to simulate besides the animation on the surface. Plus you never use raw motion capture data as it needs to be tweaked by hand.
In combination with that data you also have to simulate the movement of muscles underneath the skin to make this stuff right. If you've ever seen a model used for animals in movies you'll see that the model isn't just the outside but a full muscle and bone simulation underneath the skin as well
My take on this is that the biggest benefit of path/ray tracing is in giving indie devs and small studios the ability to produce great graphics (mostly lighting) cheaply. Rasterization-based techniques work very well, but it takes a lot of time and effort and expertise to employ them effectively. With path/ray tracing, devs just put their objects and materials and lightsources in the world and the algorithms do their magic. We'll see improvement at the top end, sure, but the big impact will be in raising the bar for the low end.
👍👍
Yeah , i'm not a game dev but i have used tools like Unreal engine and Nvidia omniverse ( mostly isaac sym for robot simulation ) and they're a god send for how easy it is to have good looking ray traced lighting . Ofc optimization is still required and needs a lot of tuning but that isn't a big issue since u have to do that in any case
Don't forget that raster lighting looks really bad in dynamic settings.
This already exists, for example Unity has plugins for screen space GI with offscreen probes and those can run even on smartphones.
@@vitordelima screens space solutions suck. They need to go.
physics, simulation and object interactivity, that's what's next.
Hoping audio path tracing and better audio-materials interactions become more prominent. Also hoping fluid sim/animations (primarily wind & water) see big refinements. The fluids animations in Avatar stood out as being extremely simplistic relative to the game's geometry and lighting fidelity.
Physics/animation systems to improve model interactions and reduce/eliminate clipping/sliding would be great
3:27 Definitely an important to point to remember as well.
And rendering subdivision surfaces (NURBs, Bezier, Catmull-Clark, ...) and displacement mapping directly in order to improve LoD, memory usage and spatial subdivision would be awesome, either via adaptive tessellation or raycasting. This is somewhat related to simulations since those can also benefit from the use of objects not based on triangular meshes.
Show me a game that REALISTICALLY portrays a character actually DRINKING from a CLEAR MUG! And we're pushing 4K res/textures running at 120fps, but nearly every game still seems to suffer from CLIPPING/COLLISION issues. Also, ANIMATION still seems to have a LONG way to go until characters consistently appear to be properly interacting with all elements of their 'environments'...
Ghost of tsushima. 4k on ps5.
yakuza games bro lol
@@superboy9431 Do they manage to show a character realistically drinking from a clear glass?
@@HiN00bs12 Does it actually have consistently 'perfect' collision/clipping and interaction with ALL the foliage in the environments? I know at least some of the plants appear to react to characters moving through ground vegetation, but most games seem to be inconsistent with it at best. Guess I'll have to take a closer look at footage of Tsushima! I still remember being amazed when the FIRST Assassins' Creed footage showed him actually using his hands to push through crowds, but haven't noticed things progress very far past that in the last decade...
@@StreetPreacherr
ye, in the first link below, others are also instances of really cool food animations.
ua-cam.com/video/t5csS7xcEzc/v-deo.html&ab_channel=ThePeteriarchy
ua-cam.com/video/_dOPdLmucJg/v-deo.html&ab_channel=BrendonCasler
ua-cam.com/video/MHYs18YRDdY/v-deo.html&ab_channel=dacyl
Absolutely! The first thing that comes to mind is the ubiquitous use of advanced physics simulations. Remember The Witcher 3's Hairworks on Geralt and monsters? Or Batman Arkham Knight's interactable smoke, paper, and rain from Nvidia Gameworks? That stuff seems to have fallen to the wayside, probably due to being computationally expensive, but it looked awesome. Both TW3 and AK runs great now on modern GPUs with everything maxed out at 4k60+ on a 30/40 series card.
Imagine a game where literally every visible item of cloth is actually a cloth sim that deforms accurately according to the geometry touching it, and all the traditional sprite based particle effects were replaced with a Gameworks style physics simulation. I think that'd go a long way to ramping up immersion.
I believe that path tracing in its current incarnation has plenty of room to grow. We don't even have lighting caustics yet where light refracts through transparent glass and water, causing it to scatter and collect in ways based on the medium of the geometry and the geometry's makeup.
But, as soon as you have every lighting phenomenon that the human eye can observe be systematic and real time, the only limitation to the realism a game could have would be the model and material assets itself, and how they behave.
I would also like to bring up, that the question itself in the videos title is kind of loaded because it makes the implicit assumption that realistic graphics means better graphics.
HiFi Rush is not a game that I would consider looks realistic, but I think it has great graphics because it has visual technologies that allow itself to look like it was an anime or a cartoon
Edit: just got to the part where john mentions this! Nice
The one thing imo that could truly make today's games feel outdated in the future is breakthrough advancements in NPC AI. While it's not really a graphics factor, it is certainly an aspect of what you see happening on screen. Personally I've always found NPC's just standing around like lifeless mannequins repeating the same lines over and over very immersion breaking.
Play Doom Eternal on Ultra Nigtmare to see real enemy AI
Check out skyrim vr gpt ai mod videos it's insane
You mean to say your AI companion running back and forth doing nothing at all is immersion breaking? The audacity
Lol
I remember seeing the original 1993 Doom, and thinking "Holy shit! Game graphics can't ever get any more realistic than this!"
One day, the games will look like the cinematics!
I remember seeing the title screen for The Revenge of Shinobi on Sega Megadrive and thinking it looked real and cant get any better.
Some games already do@@sodapopinksi667
@@sodapopinksi667I'd argue they already do
@@Alpha-kt4yl they definitely do. Way past that, even. I just remember saying that as a kid.
A bit surprised few people seem to mention artistic input. Sure, there are the purely technical aspects, but the general community doesn’t seem to understand where the artist and developer input make the difference.
Have a look at the varying results offline rendering for architecture has brought us for at least 15 years. Archviz depended on path tracing almost exclusively for a long time
Good Art Direction > Photo realistic gfx.
Also, it's pathetic how current gem games still can't handle mirrors.
Maybe it is just me but I believe more progress should be made in animation.
It can't be that we are approaching photorealistic graphics and yet characters in videogames still move in a very unnatural, unrealistic way
RoboCop is AA game I am Polish so I was watching Interview with Teyon developer, he said animations of characters looks like that because of the costs. Smaller company’s doesn’t have such huge budgets.
Graphics have a long way to go. Even if path tracing solved all lighting issues(which it doesn't,but its close) we still have to solve all the others ,like texture quality, assets quality, physics and animations. This is the future, I think with the adoption of 4K ,textures have improved vastly and will continue to improve in the future. Physics however are one place where even Alan Wake fails miserably, this is going to take several decades to perfect.
Nope. Only till the end of this generation. The only thing that will need to improve are CPU’s with physics. Graphics technology have already peaked, we just need the budgets to show us good looking graphics.
And the graphics tech I’m talking about is virtualized geometry. It basically means graphics can be infinite in detail. It’s already been fully implemented in UE5 and we just need to wait for other companies to implement the same tech in their own engines.
It these goobers in the video are to stubborn to acknowledge the potential of this tech because it completely contradicts like %100 of what they say in every god damn video they’ve made in the past few years lmao. And it also means that consoles have the possibility to literally outperform PCs.
Honestly, I am okay with how path tracing looks like now but what takes the immersion away is smoke, water and cloth physics
I really hope the next push is for real time fluid and cloth simulation
Water physics won't have a solution for a long time. Fluid physics are an especially complex area to compute and to get right without eating up all the resources you have
Path tracing is the least of it, graphically. Sure, lighting is amazing, but what about everything else in the visual pipeline. We still have a way to go.
improvement to real time physics is something I want to see after ray tracing becomes normalized
Check out Exanima, it has some solid physics in a custom engine.
Art direction is big. I was looking at the BioShock infinite DLCs, you can tell it's old game, but it still looks great. It's because of the art direction.
Also worth looking at PBR / rendering discussions that happen at Siggraph. Even some of the multi layer translucent scattering work that was being chatted about 10 years ago at that conference will be doable in tech like Substrate in Unreal in the future, but right now an extra material layer that interacts properly can add about 4-8ms on a 2080ti for a full screen asset.
But at any rate, current day siggraph film talks are always an interesting look forward for games :)
Outside of more complex materials:
We still have harsh limits on number of shadow casting lights.
Transparency is still expensive and not fully featured in most engines.
Simulation for visual effects (volumetrics sure, but even water sims, remeshing, etc) is still very expensive.
We use lights and textures that have 3 channels (RGB) instead of a channel per wavelength, or some form of spectral rendering (like in the Manuka renderer). We're probably a decade off having software we could even use to make that content.
No doubt there will be plenty of big interesting leaps like path tracing, but there are 1000s of small things we work around at the moment that will benefit from better hardware :)
There is a similar event for realtime graphics but I don't recall its name.
@@vitordelima Hmmm GDC maybe? (game developers conference?)
Siggraph does have a lot of real-time work these days too! The "Real-time live" sessions are great, often have talks from big game studios, really interesting tech stuff, etc :)
@@geoffeepIt's also part of the same event: SIGGRAPH Advances in Real-Time Rendering.
Harsh limits on the number o shadow casting lights is an issue solely caused by rasterization. The more lights you need to compute to calculate shadows the heavier it gets. Ray tracing and path tracing do not have this light source limit
Spectral rendering is definitely not needed in games
More detail for reflections and shadows, can also improve the scale of things like having thousands of super-detailed models in one scene
The more realistic games start to look, the more boring they seem. I want games looking surreal and/or 'like games'.
I agree that as graphics become more complex, it is more and more difficult to make everything coherent and fit together. When you raise the quality of one asset, you must raise the quality of everything. Making retro-DOS-looking-games is fair easier to do in that way. Those games are much simpler in graphics and gameplay hence the reason you also don't really see any bugs.
I think back in the day when 3d games became a thing, the graphical quality went way downhill. Old 3d games didn't age well at all, whereas old 2d games did.
Photorealism shouldn’t be the end goal, graphics are already getting in the way of good design, gameplay and workload for the devs.
Art design is what makes games look good. Path tracing is just like any other tool that needs to be used properly to generate an aesthetically pleasing look. Better tools don't always mean better graphics. Case in point, modern Hollywood has way better tools for CGI, but most movies from 2010 until the present look so incredibly fake compared to movies from 1990 - 2009. The original Jurassic Park had primitive tools for CGI, and it still looks more realistic than any of the Jurassic World movies.
I think one point where computing in general still has a lot of room for improvement is real-time fluid/particle simulation.
Things like waterfalls, clouds, and destruction in general still feels very "that's a video-game", Even with other visual aspects becoming more realistic every day
You can find the answer in 3D animated movies and see what has been accomplished over the last (almost) 30 years, rather than AI generated games. The main, and most important, thing remains: the artistic style.
PS AI generated games aren't new, I remember Sentdex having shown what how an AI represented GTA 5. But I'm not sure that what he showed was realtime or not.
And that has been YEARS ago.
It's nice when father talks with his sons with such a respect and patience.
Automation of processes, modular plug and play of assets and using generative A.I. to apply itself to visuals will move things forwards. But at the end of the day, it's the story its the playable mechanics that are going to define games. It's not the case for visuals any more to determine how good a game is.
For me its colour and repeated textures and lack of variation that need working on in games.
The next algorithm to get working in realtime is Spectral Pathtracing. But having enough rays per pixel to drop “denoising” is gonna be the biggest hurdle!
Thats when realtime converges with a Weta shot (strictly wrt lighting - accurate af muscle/fasciae + cloth simulation is a whole other thing - Also we have yet to see UE5.3 VDB implemented in an existing AAA title)
Maybe it’s just me but when given the option to disable ray tracing, I do. Not for the performance benefits, I just don’t like how it looks. And this is true for every game that offers it. I just feel the image is more satisfying to look at without it. Brighter and more colorful.
There’s another part of me that says I can go outside and witness real time ray tracing. I enter a video game world to escape reality.
I also feel this emphasis on implementing RT is forcing devs into utilizing “hot tech” rather than focusing on interesting game worlds. It makes me cringe when I hear the phrase “but it lacks any sort of ray tracing”.
In order for things to really advance, developers need to cut out some workloads. So shift entirely into Ray Tracing, leave baked lighting in the past that way they are not doing 2x the work to light scenes for two different types of lighting...its just a waste. Same with Mesh Shaders, leave Vertex in the dust and move on. Gamers WILL buy new hardware if forced to. But we can't move on if we are still doing "hybrid" development for a huge array of hardware, which I can only imagine adds to the cost too. Not everyone gets to have cake and eat it at the same time as everyone else.
graphics look like real-life but animations still looks like 20 years ago, we need more active ragdoll animations (euphoria engine) in games, like gta 4, 5 or red dead
Yeah it's hard to imagine where Graphics tech goes from here but logically Ray Tracing is the mathematical accuracy of the behavior of light the same could be done for all aspects of the physical realm physics material and liquid fog smoke and fire could all get the equivalent accuracy
I think at a point in the near future (5-7 years, probably before any kind of major graphical improvement beyond path tracing), games won't have traditional graphic engines and the constraints of hardware or programming. I think A.I. will be producing the graphics in real time in games and movies. A.I. artwork and animation is improving literally weekly now. Things are getting creepy; just from the perspective of how fast certain aspects of A.I. are accelerating.
Realistic physics and destructible environments. There was a kind of niche period when people were trying things with PhysX on GPUs in the late 2000s, but it didn't catch up en masse. So after path tracing becomes kind of norm in the late 2020s, just as dynamic shadows did after the early 2000, and SSAO after the late 2000s, it would probably be physics. Maybe they'll add a separate PPU (physics processing unit) onto the chip, or just let tensor cores procedurally generate damages/interactions.
Because textures are basically already up there, maybe need just a tiny bit of polish when 24~32+GB cards become the norm instead of 8~12GB.
Also, draw distance. It's difficult to notice in motion, as you just play immersively, but when you slow downd a notch and start to look for it, it's easy to notice in every game with large open locations, like Cyberpunk for example.
Actual supersampling. No matter how much you tweak temporal algorithms / DLSS / XeSS / FSR, if I have the performance of a hypothetical RX 14880XTX to 4x4 SSAA on top of a 3840x2160, I will choose it any day.
Textures, assets, animations, physics, etc. Can definitely improve. So, even if we perfected lighting and reflections, we still need to work on that stuff.
On the graphics side I want better animations, real hair strands, realistic expressions, more geometry, and honestly more lights.
By more lights i don't mean adding them for no reason, but light shafts that go through cracks in the wall. I'd also love more physics based effects like smoke, fire, water, etc. Light interacting with this would be great.
My benchmark for graphics is "what would someone who doesn't play videogames instantly pick out?". Sure, we've got traced rays coming out the wazoo and high poly characters models with sub-surface scattering. But at the same time, we've still got issues with accessories clipping into characters and the ground is just a flat texture with some 2D cut-outs of grass scattered around in tufts.
Currently a lot of image is still rasterized and old tricks are still used.
Most of the volumetrics and transparent surfaces will need to get RT to fix a lot of the problems.
Would love to see water simulation with proper refraction and IoR for next Subnautica etc.
Simulating the camera and lense system for shooting rays from lense and sending rays at different points of time would solve depth of field and motion blur.
Honestly as a 3D artist we'd love it if we no longer needed to optimize a low poly mesh. It would hasten the workflow quite considerably. I'd imagine as CPU and GPU get faster we won't need normal maps anymore for those smaller details.
We'll eventually figure all those things out. I'm more interested in if we'll eventually move on from polygons as the main medium, just like we moved from 2D to 3D 30 years ago. Something like micro voxel rendering at a high enough level would be really interesting to see, especially for physics simulations. Or are polygons along with all the smoke and mirrors we invented just the defacto most optimal way to do things by now.
I think next improvment must be in physics department. Developers just improve image and they forget physics , nut they for me is more important. Zelda TOTK prove that.
I think realistic visuals are just a stepping point along the way, stylized or hyper-real rendering will be the next step. Games where the visuals are driven by the style of the game and not by real life always have a leg up already imo, so just aiming for photorealism isn't an endpoint.
Path tracing and Photogrammetry will because close to standard by the end of this decade. Physics and animation need a HUGE upgrade and I think VR will also be next in terms of gaming but at least 15-20 years before that’s standard.
Once the paint is perfect, it's up to the painter.
'Painters' and actual painters have been getting worse and worse over time. People are going backwards, especially as technology 'progresses/advances'
We still have much distance to cover from being able to do current real time ray tracing in 60+ fps to achieving quality of what would currently take render farms days per frame to produce.
Once virtualized geometry and full path tracing is common, further improvements will still occur, but probably with diminishing returns. Real time graphics will then improve similar to how pre-rendered CGI has improved in the past years: At a much slower pace. Then the main thing that will influence the graphical result will be the art style and the quality of the game assets. The latter will largely depend on how many artists are working on them. Just like in CGI movies today.
Nanite has recently been fully implemented in UE5. And it’s only a matter of time before other companies do the same with their own game engines. And with virtualized geometry bringing the capability of infinite details, it entirely depends on budget know for better graphics.
@@balloonb0y677 Yeah, though even UE5 supports virtualized geometry only partially for animated meshes. Other engines currently don't have it at all. They will probably get to it in the next few years.
@@cube2fox no I’m the recent update the added it for landscapes. Basically every texture now has nanite.
@@balloonb0y677 I meant it doesn't fully support animated meshes.
@@cube2fox what’s animated meshes
Oliver’s voice is so smooth. I want him to narrate all the things!
Graphics were good enough five years ago, and they are especially good enough now.
Make animation better. Make physics better. Make gameplay better. Do stuff beyond “look at these shiny objects that are in a world of bugs and glitches”.
Make gaming great again.
Very true, but as time goes on it is all about the profits. Making games to make money, instead of making money to make games(with passion).
At best current games look like realistic cartoons, not realistic realism. Long ways to go.
The next thing after path tracing is wave tracing. Light in the real world has properties of both a wave and a particle, while standard path tracing only considers the particle aspect.
The same question was asked before path tracing was discovered. And before physics were introduced. And before shaders appeared. And before 3D took over.
Etc, etc.....
To me Graphics improved when RT and Motion blur is turned off .
Nobody is choosing 30 fps with RT over 60fps without RT.
Shiny floors are not that important .
You really don't understand what RT does then. It's not just shiny floors that it does. It's the entire lighting pipeline that is being improved upon. So what it's expensive, other visual effects like shadow maps are still expensive to this day
I believe that once they incorporate the new AI dl models for NPC personalities, combat and the like this will overshadow ray tracing completely. Graphics are sort of at their peak and in this day and age games need something more than visuals. There are many games with breathtaking graphics but lackluster gameplay. We need something new, not 'lighting affects'. Just doesn't cut it by any means.
This question whether we peaked is a failure of imagination.
Yes. There is always room for improvement. As long as I can wait 10 hours for Houdini to finish a 10 second fluid particle simulation, as long as you cannot simulate the sahara with every grain of sand reacting physically correct, there will always be room for improvement. Because you can bet your butt if a triple A gamedeveloper had the resources to render every grain of sand in a desert, he would totally do it. As long as a high detailled VDB voxelcache of a nuclear blast can easily take a TB of cache space, there is things that are still years away from working in realtime. Ask those VFX guys in Hollywood about caching times and rendertimes. When it comes to complex destruction simulation with true to world detail, even seconds of animation can take up hours and days of caching time and literally Terabytes of space.
Of course AI will speed up things a lot but only because you can RENDER everything, doesn´t mean you can get everything imaginable to work smoothly in realtime. Let´s just look at water. A closer look. Or think of fire. Chemical reactions like acids. Not even getting started with caustics, dispersion, diffraction, complex translucent materials. There will always be a fraction of users longing for the most immersive and realistic experience and there will always be a tech company looking for reasons to sell you an even better product.
This is a ridiculous notion and a failure of imagination. All ConcreteLlama had to do is go look at things like the featured shorts in Love, Death & Robots, or any number of celebrated animated films to show just how FAR realtime rendering has to go. CG movies have had pathtracing for ages, and yet they still continue to improve dramatically.
For games, the focus should be on dynamism and interfaction. Make it so you can interact with practically every element in the environment. Build it up, bring it down, exploit natural forces, be subject to the effect of natural forces, all while robust physics engines and robust pathtracing setups support the entire effort.
Path tracing has a long way to go. Most games only calculate the surface level of lighting while complex materials for skin, cloth, metal, liquids, glass, fog etc are still a simplified approximation. There are good reasons rendering packages like 3dsMax, Maya, Blender still take more than 1h per frame without any upscale or denoising trickery.
I think when all games look super realistic, it'll be hard to differentiate what game it is just by looking at it and we will start to see more focus on the art style instead
The funny thing is that on consoles I miss realistic looking and realistic playing shooters. For me, they all have too much gimmicks going on, like the extra powers, more focus on style, skill trees, wall jumping, battle royale, etc. I wish we had more games like Brothers in Arms of Ghost Recon kinda games. But hey, I'm getting old I guess.
All games won't go for a super realistic style though. Path tracing doesn't mean you have to go to realism for your visuals. Plenty of games already focus on different styles besides realism and that isn't going to change. Obviously not every developer wants to aim for realism and that hasn't changed at all
dammn so this means my laptop will have to be replaced when all the games start using it
Gaming Laptops don’t typically last very long sadly
@@pistachiodisguisey911'Gaming laptop' is an oxymoron.
Alan Wake 2 looks great but it would take 1 second for mostly everyone to identify that it's a game. We are very far from photorealism. Even movie studio with rendering farms often struggle to reach it.
When we look at it component wise I think our GPUS have started to reach the limit with path tracing, and I think for better realism our next target component wise is ram and cpu. We need real time simulated things now.
I still remember standing in front of a ps1 kios playing destruction derby and thinking to my self, this is it, i dont ever need better graphics. XD
Geometry, textures, physics simulation... yes, a lot can be improved.
Physics, easily. I get kind of upset by people being so oblivious/inside the box in regards to video game visuals, means that devs will continue to upgrade what doesn't matter (textures, geometry etc.).
After that, I think video game tech is officially completed. For the worlds to look photoreal and ALSO you can interact with it realistically on every level? Water, hair, cloth, destruction/deformation etc. That's the endpoint. And then again, that's an end point faaaar off. So don't worry, it's gonna take a while.
When we've hit perfect light and physics in games, and the assets/materials are indistinguishable from reality, thats when skills from movie production should transfer over, we'll see more cinematographer and set designer style roles. It'll be very freeing and much like painting, when photorealism has been achieved the next step is moving into more surreal explorations of design. interesting times ahead.
No it hasn’t. Look at pre rendered trailers or demos and you’ll see the differences from a real time render.
i still think work needs to be done to character models faces still they dont look real, you know its a computer image compared to the clothes there wearing were metal and leather looks almost real
you seen games like unrecord that are going for super realistic enviroments hide faces because they no it will blow the realism, its just not there yet
Character faces is a combination of both lighting and texturing. You need to simulate or at least fake the way light scatters when it hits skin which is called subsurface scattering. You also need to make sure the texture work shows the necessary detail to make it look right
Crysis from 2007 was a good start to faces. Very impressive for that time. Even now they are better than basically all games.
i noticed that when u lower the pixels like going from 1080 to 480p in gow 2 makes it look so fk reallystic even more than games like sekiro in 4k because in 4k u can see every little detail and every error they did when making the textures
Path tracing in games still not fully accurate, we need 100x more powerful hardware for things like colors from oil spill/CDs/prism, frosted glass or any material that uses light's wave property, movies resolve it by pushing millions of rays of different frequency per-pixel.
yes, tons of stuff...ultra realistic body and face for every npcs in the game world, realistic non robotic animations for everything, physics...seems gta6 will be the next big improvement in those areas according to the trailer.
Man .. I used to think half life 2 , Max Payne 3 etc had the greatest graphics ... Because the games were imeraive and good ...and our minds interpolated it into super path tracing... But I play modern day games and
Remember folks. Path tracing is a tool. It's how they use the tool that is always going to get better and better.
Also Path tracing is just the beginning. Light is not a ray scattering around. Light is a wave and each color has its own wave length and is also affected by the physics like gravity as well as interaction with the infinite amount of wavelengths being scattered around like sound waves.
There is so much to improve on we are just not even scratching the surface yet. But we just FINALLY! started to attempt to scratch the surface.
Developers will not spend weeks and months trying to fake the lighting with raster based techniques. They will actually spend weeks and months thinking of how to better bring the lighting to their intent instead of trying to imitate what was their intent. Big change there.
No more *"Ghost towns."*
Finally have high density, dynamic and diverse NPC populations.
Until you can't tell if you're watching a movie or a video game there will be room for improvement in graphics.
CIG just demonstrated very good water physics, fire propagation, and wind effects with their Star Engine. When you see how many super immersive features they've added to that engine, the long development time makes perfect sense. Bethesda could never. 😂
I think it’s more cpu side stuff, more complex ai in the npc’s, more npc’s more simulation and physics for dynamic intrinsic gameplay etc than more shiny stuff that’ll feel next gen, rdr2 wasn’t just pretty (without ray tracing too) but had a complex world with a lot of moving parts where cyberpunk is very stiff and sterile beyond visuals
still a lot to do in the hair and cloths department. Bending metal armour still breaks my illusion
I feel like clipping as well as hair are the final frontiers of graphics.
Of course, things like geometry and texture can always get better. And path traced lighting still can't recreate how reality looks, only try to approximate. Put it this way, if graphics can't recreate what you can see right now with your own two eyes, then graphics still has a long way to go.
Imo one of the future step will be 3d or "4d" gaming
FG is good when it works, if they reduce imput lag it is ok, if you interpolate frames and i have 133/200 ms rendering time frame : i dont care ...
I wonder what these guys would look like if they got jacked/ shredded
Obviously yes. Movie VFX have been path traced for over 20 years and they are clearly (as a trend, not every example) getting wildly better year over year.
Devs need to stop obsessing with vram destorying details and focus on leveraging cutting edge technology to bring superior art design to life. Look at rage for example. Cutting edge tech brought a style of graphics way beyond its time. Fully baked in raytrace lighting. Yes static but a great first attempt
Before pathtracing the only thing getting better in graphics was the resolution. That is because the ps4 and xbox one held back the generation. Today you need at rtx 4090 to max out the newest games but in just 5 years i bet you that will be considered the base line. Whats next in computer graphics is close to photorealism but i think thats far ahead. Future games will probably be a fully functioning reality just like our reality we will be able to simulate reality inside our own simulated reality and the end goal is to escape the boundaries of our reality and enter other realitys to explore.
No one will care about graphics anymore when graphics is photo realistic.
Path tracing is just lighting. There is way more to game graphics than just lighting.
Was expecting some mention of realtime simulation, translucency, subsurface scattering, mocap data with procedural generated in between, morphing. Gaming is far from cg in terms of graphical details because of optimisation. A simple comparison between both will list a lot…
The graphics don't need to get better - path/ray tracing will make the current gen graphics look like real life (and more importantly, the games need to be more fun)
Enough with shiny textures and whatever lighting. Give me true physics simulation and destruction. Give me what Crackdown 3 originally promised.
True physics simulations aren't possible in real time
Try Exanima, it is full of physics.
I’m just waiting for path traced Minesweeper
I would play that
I still think ray tracing is overrated. You can fake lighting and still make it look realistic. I'd rather see realistic details density, realistic environment physics and reaction to movements and realistic animations especially humans.
i asked ai it said something about hybrid rendering-.
Bidirectional path tracing is the next big thing
Is there a more advanced light transport technique than path tracing or is path tracing the end game?
Lmao we're not even close to offline rendering quality. Socalled pathtraced games are still very limited.