Why Does Naughty Dog's 3D Environment Art Looks Like This?
HTML-код
- Опубликовано: 7 сен 2024
- In the previous video we looked at what good optimization for 3D game resolution assets looks like. This time, we're looking at Environment Art topology, specifically from Naughty Dog's Last of Us game and discussing why it looks the way it does.
Mentorships | emc3dart.gumro...
Portfolio | www.artstation...
Twitter | x.com/EMC3D
More examples from the mini lecture can be found at:
EMC3D/highlights
Artwork used as examples in this video:
Edgar A.Martinez: artstation.com/edgar_a_martinez
You wanted a longer discussion/vibe about topology? Well let's get right to it then 😃!
ruclips.net/video/0gU2VxsdB1E/видео.html
You made a really important point.
The key is vertex-based texture blending.
Assigning a single material to an object in the usual way would require creating a large image texture for the entire house or adding tons of decals.
For example, in order to display mossy wood with high resolution on each wall of a house, each wall may need to have a texture. Otherwise, you'll have to put countless moss decals on every wooden wall. There are also problems that may be revealed by repetitive tiling.
However, if you do vertex-based blending, it can be easily done with a few materials. The blended area does not repeat and can be applied differently to each object area.
Vertex-based blending does not reveal pixel artifacts, unlike texture mask-based blending.
Complex UV and Trim settings can be reduced for precise texturing.
Naughty Dog's method of creating backgrounds increases the polygons of the assets, but this greatly increases the texturing benefits.
Let me take this a step further in a slightly different direction. Many games also use a technique called material layering. This is not unlike vertex blending, but instead relies on grunge/ noise maps to mask out between different materials. This means the environment texturing process becomes a bit like creating materials in substance painter.
So, you could optimise the models to have huge quad faces and optimised triangles in just the corners to significantly reduce polygons in a scene and spend the budget on blending UV mapped or tri-planar materials together using blend masks and then place certain types of details like stagnant and rotted drip stains from old window sills using Decals or better yet, mesh decals :)
I went quite deep into a while ago when trying to find the definitive ways of doing things only to discover the techniques change based on the company's pipeline or requirement.
Wouldn't surprise me in naughty dog also used blend materials.
A really obvious and prevalent use of this technique can be found in Elden Ring and CyberPunk 2077.
Edit: nvm I just re-read your last part of the comment and noticed you already mentioned this. I'll just leave this here anyway so it helps explain what it is to someone who doesn't know.
With some work texture-based blends can also avoid the pixellated look. I think vertex-based stuff is preferred, though, simply because it's easier to edit on the fly and takes up considerably less memory.
The even, uniform quads also help with meshlet culling which is huge for opaque static mesh performance on modern hardware.
Wow, I learned a few things from a 1 minute video. The last of us is the game that inspired me to start learning 3D software and it's great to see how the artists at Naughty dog approached it. I would love to see longer video like that, explaining even more.
fucking damn beautiful topology
How is this a good topology? It clearly has too many loops.
@@patek2385 I don't know if its actually the correct way to do topology, I just said it looked beautiful i'm not a 3d modeler expert either, even this is better topology than my model's
@@R4v3nB14ck good topology would be if the straight walls had no loops and less or none vertices. Windows would be connected via 4 lines only (on each corner). We wouldn't see any quads whatsoever.
The topology shown in the video would be great if it was inside 3d modeling software only.
@@patek2385 But think about it, they actually need those loops for the Game Engine, do you think It could have been made better? I mean maybe there are not bad topologies if you have a reason to do it that way, just like music
@@R4v3nB14ck Probably they had no choice, limitations of the game engine.
Love this explanation! Industry intel are very rate to find on youtube.
Cheers chap!
I watched the GDC video of the naughty dog artists that worked on the vegetation for last of us pt 2 remaster. I learned how they used vertex painting and now I'm using the same technique in my game
I still don’t quite understand what you meant by blending the texture more evenly. Can you make another video explaining it?
Basically, say you have 2 textures, brick and moss. You start my covering the mesh in bricks, then to make your mesh more unique you want to paint some moss on it. There are a few techniques to do this but the most common is to do vertex painting, which is where you assign some weight to each vertex to use the moss texture. You then get some blend between vertices assigned different weights of moss. The more vertices you have the more control you will have over the transition between the two materials/textures
Think of the mesh grid like pixels in a texture - you need decent resolution to see your brush strokes
@@duke222222 I’ve made tons of props in my career and I can still texture everything well, as long as the UV is good, it doesn’t need the topology to be that dense. Some games even require as low polycount as possible for optimization, and I can’t afford to add polyloops where it’s not necessary
I understand, but this is specifically for vertex painting that relies on vertex density. I’m not saying it’s good or bad, just what to compare it to.
@@duke222222 I think if you have denser geo, you can use more accurate vertex paint as ID mask, so that makes sense
Thanks alot for all your tutorials, its like a goldmine for an game artist! One question: Do you think the walls of these houses are made with planes oder cubes? Chatgpt tells me that cubes are better because they bring a physical thickness that looks more realistic.
Hi Faryan, I will answer this in an upcoming video/short, as it's a great question!
@@EMC3D You are awesome!!
thanks for the quick video
No problem.
Man, I remember a few years ago that I couldn’t understand this stuff, and now it’s like I know a second language that I’m at least decent at and developing further!
That's awesome, a great sign of your skill progression that just gradually happens as you keep doing it.
@@EMC3D thanks! Right now I’m learning Maya…it’s true what they say when they say it hates you. 🤣
That and it screws you a bit when you look at how Blender’s controls works. But it’s a learning curve.
This videos are way too cool for how short they are.
When you have the time, try to make something a bit more complex ^^
Is the performance loss worth it, considering that you add so many unnecessary vertices? Wouldn't using other methods be a lot better for performance?
Absolutely yes, if you consider the memory saved by using trim sheets.
It's all a balancing game indeed. Thanks both.
This is way too short to be any kind of useful in any meaningful way.
You're either not listening or already know why ND did what they did
Perhaps you should learn some basic concepts and come back later
I think I can make it even shorter than that. Basically they are using subdivided geometry to be able to do some pretty detailed vertex painting, which allows them to add unique details to each object (or each instance) and make things look more interesting.
It really is a matter of knowing basic concepts when trying to understand a quick showcase like this.
Cheers chap!
What is the need to add subdivisions in flat objects. When they can be made with only 4 vertices
Vertex painting
Cheers chaps, yep and a few other things folks have mentioned in comments, vertex lighting, baking info into verts, trim mapping, etc.