The following video shows how gta could look in ue4, and although its only 1 mans work it doesnt make unreal look good. Is this because the artist hasn’t created next gen looking terrain with world machine or custom foliage materials with proper lighting & no properly textured/baked assets or decent materials ?
Can UE4 actually look as good as gta5 & perform just as good too if done by proper artists ? Budget aside, when I look at this guys work I know many of us 3d artists can do better, at least I could do much better in the foliage & terrain department, I just hope I’m not limited by the engine
I do see some things that set it apart from rockstars engine in the lighting & shader / material department, is this the artists novice skill or ue4s limitations?
Sure, UE4 is capable, the deal is though you’d need millions of dollars and hundreds of people working on it. The work of one guy is not a representation of that capability. What they were able to pull off with GTA5 is extremely complex in how they were able to manage all the complexity and maintain performance.
If you have to ask this question with that comparison, then no, UE4 will not limit you in any way.
I’m not trying to be rude by any means, but Rockstar had a lot of people and a lot of money. Is UE4 capable under those circumstances? Yeah. If you have those circumstances then you wouldn’t be here asking us.
Ultimately, UE4’s performance and quality is incredible. I take that’s the answer you wanted.
UE4 handles lighting a bit differently than GTA V. GTA V stuck to the old model from UE3 which used light environments, phong shading, and forward rendering. Each dynamic light added multiplies the difficulty of rendering the material, so 10 lights shining on a surface would give you **** results. UE4 uses new algorithms to calculate lighting and complex specular reflections. It’s a lot cheaper to get per-pixel reflections in UE4, but it’s also a hell of a lot more expensive to get it more accurate. If you don’t mind the accuracy suffering, you can use generic cubemaps and no reflections to handle your lighting. Mario Kart 8 basically did that and got PBR to run on the Wii U, which is not much more powerful than an integrated graphics card on a laptop.
There are new arrays for handling texture samples in 4.6, and much more efficient instancing in 4.7, so if you wanted to make a huge detailed game like GTA: V in UE4, not only would it look better, but it would run a hell of a lot better and be easier to work with, too.
It’s not so easy to say that it would run better on the same level of hardware. There’s going to also be alot of CPU based bespoke code and optimizations that GTA has in order to run a world full of dynamic objects that Epic don’t have.
And we don’t know how well Euphoria runs when pitted against PhysX, although I wouldn’t be surprised if Euphoria is a much more efficient engine since NVidia like to cripple Console and AMD owners.
I always thought that UE4 is losing its AAA vision and turning into indie.
It looks not bad for an indie game but ridicilously bad compared with GTA5.
I think the fault does not belong to artist, I am pretty sure there was 1 maybe 2 max 3 artists for this project for very limited time.
Do you know how much artists are working to design even simplest scenes in Rockstar ?
Every engine has upside and downside compared with other engines.
Let’s check downsides of UE4 graphics side
-No 8192 textures
-Light calculation times are still way too much and I am pretty sure it is not using any GPU power to calculate, that is why people are still using even 64x64 lightmaps, this one could be the weakest side.
-Dynamic and even static lights are not well with general accuracy.(worse than UDK)
-Quality and speed of screen space reflections and screen space ambient occlusions are not satisfying.
-Texture compression times are a little much than expected.
-Lights are just 2D textures, none of them gives 3D volumetric effects, especially spot lights are ridiciliously flat.
-So much sacrifice to adopt deferred shading.
-The editor is slow that kills enthusiasm to try something new quickly.
-Texture streaming is not working and just pain with standart hard disks.
-Blueprints editor is also slow.
I think the lowest factor is lightmap resolutions. Even you may use 1024x1024 lightmaps but because of wasted space on UV you probably get pixel resolution of 512x512.
The long calculation times does not allow lighting artists to try something new and see quick changes. It just kills excitement.
You may try something new with dynamic lights and then when you are satisfied you may convert it static but it gives %100 totally different results, even lighting intensity may have huge differences.
Lighting is an issue, but in the long run I suspect it’s going to be the lack of fully multithreaded game code that’s going to hold the engine back when compared to what Ubi Soft, Rockstar & EA have up their sleeve - that’s really going to help with dynamic, densely populated worlds and fast streaming.
And as for AAA Devs, AAA Devs were heavily customising UE3 throughout its lifespan. and it’ss going to be pretty easy for big devs to add their own lighting solutions pretty quickly now that PBR is standardized across the industry.
If you use your assets properly, you can remove the tiling look on landscape. 8192 textures are heavy, difficult for artists to make, and totally not necessary. And UE4’s lighting is really good for what it’s trying to do. I wish it had sharp caustic reflections and more effective transperancy modulation built-in, but for a game it’s seriously awesome. Just build good shaders, use good techniques for texturing assets, design your spaces wisely with the assets you have, use all the tools you’re given, and make something awesome. I find it so sad that people still plug a texture to diffuse in the material and call it “done.” Then they look at the finished product and complain why it doesn’t look as good as CGI. It can take me hours to work on one shader and get it just right.
SSR reflections are actually very good in UE4. The problem is the reflections are SCREENspace, so if your object is trying to reflect something that’s not visible on the screen, the reflection value returned is 0. This is why you use skylights with generic cubemap reflections to fill in reflection data that might be missing. With that implemented, everything in the scene will have an appropriate reflection. This is a limitation with the SSR method in general, NOT UE4’s implementation.
Lightmass is a general solution for GI. It’s not ever going to be perfect. When you combine that with SSAO and SSR and area lights, you can get an incredibly pleasing result. I’m not sure what you mean by lights being “2D textures.” Lights are not 2D, pointlights are calculated in all directions within a 3D space. You can turn on volumetric effects if that’s what you want. Maybe next generation when SVOGI is actually implemented we’ll get awesome, awesome perfect dynamic GI, but in the meantime, LPV works well for outdoor environments.
If one person made all that in UE4, imagine how good the game will look with 1000 professionals working on it for 5 years. That’s basically the development cycle for GTA: V. There’s a lot of awesome stuff in UE4 that people are just ignoring. You won’t get instant gratification in game development: everything good takes time for every engine. It’s not UE4’s fault that there are so many “doors” to take care of.