Good Topology: What is it, and Why Does it Matter?
Вставка
- Опубліковано 21 вер 2024
- Every time I make a video about topology, people ask the same questions. May as well sum it all up in one video. What is good topology, and why does it matter.
Support this channel and get access to exclusive content on Patreon- / decoded
Facebook - / blenderdecoded
Twitter - / decodedvfx
#Blender #DECODED #B3d
If you enjoyed this video, you might want to check out my course, the Essential Topology Guide. Use the code LAUNCH to get 20% off at Gumroad -
decoded.gumroad.com/l/ESSENTIALTOPO
The concept of "good topology" isn't just about the positions of vertices and the ways they're connected to form faces, but also the purpose and use-case of the model. In the context of subdivision surfaces, Suzanne has ok topology that's built for that purpose. She isn't built for clean deformation which is important for animation, so in that sense, she would have bad topology. She isn't built to be used in a game engine, so in that sense, she has bad topology. She could be subdivided a number of times, and one could build a low poly model over the top to capture her details, so you could say she has good topology for that. Topology is all about purpose. She could be composed entirely of tiny triangles, and that might be considered good topology if it served the purpose it was meant to.
Yep, it seems that there's so much misunderstanding about topology once you scratch the surface.
The creator goes over all of this in his video
Finally someone explained it clearly and in its simplest form. I've been watching a ton of videos about topology and no one said it better than you. Hope you can continue more videos like this!
How would people feel about a topology course in the future?
So far in my 3D education, the most difficult thing has been understanding the whole workflow end to end. There are so many differently workflows and topology requirements can be very different for each. It would be great to explain the workflow implications for each of the common topology types. For example, game dev low poly workflow, game dev character workflow, VFX deformation/explosion workflow, etc.
we need it asap,thanks for the vid
Yes please! A series tackling different model of different styles and proportions for animation would be massive!
Yes, havent been able to find a good one that starts from the bottom and explains theory to work to complex things.
Thanks for any videos on typology - I can really use it - typology to me is always a fight.
Just a small note: quads do not always have 'at least one set of parallel lines' (4:45). A quad with at least two parallel lines is a parallelogram - a special quad.
They often have in 3D modelling though, that's true.
Quads can be bent into non-parallel shapes, yes. But that usually isn't the case, at least by default.
Well, just throw in Suzanne and you'll find plenty but ok.
@@DECODEDVFX A quad is just a polygon with 4 edges, aka a quadrilateral. It's very easy to make a quadrilateral with no parallel edges out of a square. Just take one of the vertices and move it to or away from the opposite corner vertex by any amount.
typology
Key things I understand from the video is this: focus on edge flow rather than obsessing over quads, shading is more important obsessing over quads, and overall topology is something to focus on and localizing topology for detail as well as ngons and tris are okay sometimes.
That's exactly what topology is: a mathematical representation of edge flow, by drawing lines across a surface, which is traditionally done in an interval grid to create quads.
@@locinolacolino1302 In my college days, we did do drafting by hand to understand engineering drawings so we can present CAD models properly. What did we use? Grid paper.
Thank you. Useful information.👍
Glad it was helpful!
I've recently been making some character mods for Bomb Rush Cyberfunk. Stylistically the game is low poly, flat shaded. One thing I've tried to make work is pants wrinkles. Characters wear baggy-ish pants and you gotta sell that with wrinkles. You have to achieve this with basically a shaded solid color. Not easy. Understanding what I could get away with in terms of bad topology allowed me to really sell the wrinkles while making sure the rigging was cohesive.
Thankyou. I use Blender for the most basic concept design so thought this wasn't that relevant to me but it was really interesting and will alter my workflow in some cases. Brilliant.
And completely devoid of the distracting meaningless fluff of "go ahead and" that seems to have infected so many Blender tutorials so thankyou twice!
Good and bad topology sort of depends on the aplications. Quads are usually preferred because they work preditably with sub-d workflow and with any deformations like rigging/anim and displacements but render engines use triangles because triangles are planner and it's the easiest and fastest to draw for a path tracer or any render engines which why in games most models are filled with tris.
I've also been told one of the benefits of quads is that they can bend along either of their diagonals, which can make animations less prone to artifacts. My understanding this allows a rendering engine to pick the best diagonal to triangulate on at runtime, so it can take better advantage of dynamic deformations. If you're working with a triangulated mesh, your diagonals are fixed, so they may be at odds with the angle a face should deform to follow a rig.
The purpose of the model on which I work is to create a complex assembly of organic shapes for 3D printing (+printed molds); the rendering and animation are not needed.
A good topology couldn't do any harm but:
-It shouldn't be done before freezing the design.
-Would it be worth the time invested?
The topology matters only in the sense of what gets your job done the quickest.
Do not avoid N-gons and triangles, if that kind of topology gets your work done quicker. Whether the mesh is triangles, n-gon, or quads literally does not matter, if your final render looks the exact same. Only think about nice quad flows if that makes your work easier, not harder.
This is why for Blender to have the ability to do n-gon holes on n-gon faces like it is possible in SketchUp, is super important and should be a top priority.
The time benefit is mainly observed later down a studio pipeline, good topology is easier to unwrap, texture, simulate, shape-key, results in less lighting glitches. But as a one man band just go with whatever works.
@@locinolacolino1302 Depends. In most cases, the "Good topology" adds no benefit at all, especially if the model never leaves blender...
With quad topology you can easily create lods, however I think ngons are much easier to manage in order to reduce the polycount.
@@gabrielegagliardi3956 Game engines will automatically split NGons into tris upon import. Feel free to triangulate in Blender to see the model how the Game Engine does.
@@locinolacolino1302 This is best practice anyways. Always slap on your triangluate modifer. Don't let the game engine triangulate on import and screw up your vertex order. It is important for baking and to avoid discrepancies in how faces are created to have your model trinagulated BEFORE export
Heres why I wish that I think people should learn more about "correct" topology. Having the correct topology can be time consuming, but not as time consuming as troubleshooting your models, texturing, baking etc etc.. It just saves time to have good topology or at least know about it to be able to quickly fix problems.. I dont know how many months I wasted as a beginner just troubleshooting things getting weird problems that I didnt understand how to fix. All the hours wasted in Substance Painter because models looked like absolute crap.... All solved when I realised the importance of good topology. I havent seen Decodeds course and I we dont know eachother, but I think that I can safely say that from knowing the channel that I'll recommend his course to any seriuos beginner.
The vast majority of problems concern shading, because when you import the model as fbx or obj, the mesh gets automatically triangulated. Quad modelling rarely gives shading problems because a plane (face) is perfectly split into two triangles, on the other hand, vertices in ngons faces try to algoritmically reach other vertices to form triangles, basically distorting the mesh' s shading. I know you already know all that, this comment was more for the benefit of people reading this in the future. Cheeerssss
@@gabrielegagliardi3956 To prevent this, you simply triangulate your model before export, always with the same algorithm. No single quad needed just to ensure that. Shading can be easily checked and controlled right in DCC before the model goes into production.
Normals matter so much. Especially if its a Blender to Unreal Engine work flow. When I port an asset into UE5 that features an Ngon it evidently creates shading issues but if its all triangles/quads its fixed 9 times out of 10.
That's common sense. Models ALWAYS need to be trioangulated before going into any Game Engine, to ensure Vertex Order for Baking and shading. This is because if you throw a non triangulated model into any game engine or real time rendering application like Marmoset, it will run its own triangulation algorithm, which likely will differ to the one on your DDC. This has nothing to do specifically with Blender or Unreal. Applies to Unity and whatever have you, too.
I have seen some characters done in quads and the mesh looks amazing, but I never found out how exactly were they made. They look perfect like the mesh was made by a machine and not a person. There must be a workflow that is followed to produce such complicated and exact results every time. But what is it?
Try cube modeling or Retopoflow. it often is done by a machine.
Now we want a tutorial on how to make a bad mesh *that* terribly bad at 1:10
Convert to tris. Remesh. Decimate.
@@DECODEDVFX Flip a few normals and split corners while your at it 😂😂
one point to note is that if you have triangles and/or n-gons on flat surfaces that are going to be animated, and the animation involves those faces being deformed from being flat, you're going to have issues
I have a very simple set of rules. As long as an object doesn't:
1. Bend/Deform in any major way
2. Have an obscene amount of polygons
3. Take up a lot of the screen
Then it really doesn't matter how good/terrible the topology is. I mean sure, don't do this for your portfolio. But if your only goal is to get things done, then you'll probably have to accept some parts being a little crappy.
You still at the very minimum want to be able to use modifiers with pretty much any of your meshes
@@bobross9370 I'd argue that depends on the scene. The model is just there to fill out the shot? Then it really doesn't matter how well it plays with any modifier other than maybe decimate.
@@demian5631 It sounds like you may not understand why quad topology is important. It doesn't take that much more time to model with correct topology than it does to do a jank/crap job once you have the principles down and you've had some practice. The main difference? Other people can use/manipulate the model, it will work with modifiers, simulation and animations, you don't spend time fanangling or fixing/hiding errors later when things go wrong. Professionals don't really tolerate bad topology in many instances in the industry and in the end you'll understand and learn that even though your ego may think you're better than the decades of experience others have spent learning how to 3D model more efficiently, you'll eventually realize that Auto-Smooth Shading isn't your savior. Modeling and mesh principles are still the same as they were when we first started. Which means the mesh topology rules and guides are probably not obsolete! I don't want to discourage you, as I have gone through the stage where I thought I was better than I was, but it took some humility to realize that my cutting of corners was just wasting my time in the long run. Do it right and you'll do well.
@@demian5631 Pretty much what the other guy said, modelling quads is not hard at all, you could even do a lazy method on background objects with a quad remesh so long as it's workable topology. If you ever try to take 3D seriously, you will very quickly find out how important quad topology is. Modifiers matter a lot, it's best to have your options open on every model rather as struggle one at a time.
@@bobross9370 If a model really needed it, then I would load the model up in instant meshes. Gives much better results than Blender's built in remeshers.
My point wasn't that you should never care, it's knowing when to care. You are hired at a studio and have to work with others? Make sure the topology is at least decent. You are working on a project entirely by yourself with little time and no money and the object in question is out of focus and 10px big? Who cares! I've used alpha masked images as backfill before and no one has called me out on it yet.
Spent your time wisely is my point.
Awesome video that was really well-articulated thanks! Am I correct in that non-quad topology may interfere with a lot of modifiers too?
nice video , thanks .
So the reason that you should use quads over tris is because they’re easier to work with, good to know!
Thanks for this!
In 4th minute you said that quads are good because they are made with parallel lines. Ehm, there are a plenty of quads that doesn't have parallel lines, Did I miss something?
Most quads have at least two sides which are parallel or close to parallel.
When the lines are not parallel they instead follow the contours of the object usually. The lines concept still holds up even if they are not all perfectly parallel/perpendicular. I'm sure you are thinking of loops around the lips or ears on a face model or other unusual organic shapes, but the general flow of the face still has lines that flow upward and left to right, or across the contours, maintaining the continuity with the loop of faces its connected to. Correct me if I'm Wrong anywhere @DECODEDVFX
I suppose that bad topology is like poor word choice. That is, if the point is communicated, then there's no problem. Good topology communicates desirable aspects of the surface. If tris or n-gons aren't confusing the intention of the surface, then ignore them. In that regard, your example of bad topology (the triangulatized monkey) is ironically good topology because it effectively communicates bad topology; it served the desired purpose. If you switched the two examples, _both_ would have failed the desired purpose 🤔
My math teacher Ethan Concin traumatized me with this subject
i made many models for games with shit topology but the only thing that matters to me is the shading if the shading was clean no need for good topology
Well done :)
Nice video :)
Thanks!
they do not assuem the normals are facing front. from the face. its based on the vectors . that are stored with the point data to make the face. not speaking of face orientatin but the vectors them self. the face normal is the sum of it's verts
Good topology is like clean, extensible code. You can write a spaghetti monster that "works" and gets the job done, but the moment you try to extend it or change it, you kick yourself trying to de tangle all the spaghetti without breaking stuff.
This is a good example and helps me understand better. A lot of people talk about this topic from “professional” perspective, and all I really care about as someone just doodling around is “ok, but how much does it really matter?” The code example makes much more sense to me. There are times where it matters immensely, and times where it doesn’t matter at all. It depends on end goals, what you’ll use it for, does speed of execution matter, is there any reuse, are others going to need to be able to read it or modify it, is it a side project, is it a professional product, etc etc.
The most important reason triangles are used is that they are the only shape that is guaranteed to be coplanar while still having a surface.
I think what you say is important to consider and understand. I’m new to this and I was looking at a model with all quads in blender that seemed to have way more geometry than the faces would initially seem to allow. Upon closer inspection, I realized that the quad faces can, as you imply here, be 3D.
To help me visualize this I had to think about a flat square, then if I move opposing vertices in opposite directions - eg two opposing corners go up or the two other go down - that’s like folding the square into two triangles that are not co-planar. However, there’s no combination of vertex moving that will make the triangle be anything but co-planar.
@@JayM928 Precisely :)
Something else to note is that most engines convert 4+ polygons back down to tris when rendering. This isn't a problem when the quads are coplanar, but there are two ways to make two triangles from one quad and they can produce drastically different shapes leading to unpredictable behavior (never a good thing)
The only reason triangles are used is because graphics cards ONLY understand triangles and they are efficient. Nothing the fuggg else. Most game models are human readable during authoring time, they only get triangulated on export fully.
@@harrysanders818 First of all, where do you get that idea? Because its completely wrong. Second, why do you think graphics cards only understand triangles? Don't you think they would be designed with the hardware to hand quads if it were just that? Don't post comments like that when you have no understanding of 3d graphics or experience in the industry.
0:44 jumpscare
I didn't think I looked that scary.
to be honest - Suzanne eyes - is a sample of worst topology. Poles never subdivide smoothly and are very costly to compute, and using them in such visually important objects like eyes - it is a crime. But it is shown for decades, that it is ok practice, especially for newbies.
And another video that gets it wrong.
Can't enumerate the cases where I had to feedback beginner to intermediate 3d artists game res models who are damaged by this kind of content, which is actually making them less employable. They're putting themselves under immense stress to create what YT calls "Good Topology", creating nonsensical "deformable" quad structures on static inanimate objects and such. There is no "Good Topology", nor do quads matter on final game production models, especially not static ones. While this video sort of points this out, it still commits to the usual nonsensical doctrine all in all. Don't focus on topology, focus on gaining your art skills, and the tech will fall into place, as Topology has much more to do with understanding Form and Silhouette than technical tools. Don't believe the topology hype created by people who try to compensate for their lack of training or effort in doing the actual art and instead overly focus on topology and tell you it'll make a huge difference. What will make the difference is learning the fundamentals. Topology is not your actual brush stroke people will value or awe.
You mix up topology with geometry, mathematics-wise. Both Suzanne's have the same topology, but different geometry. Geometry deals with quantitative properties of space, such as distance and curvature on manifolds. Topology deals with qualitative properties of space, namely those that remain unchanged under bending and stretching (a cup with one handle has the same topology as a torus; see also Euler's formula).
Mathematics and Modeling - like any very specialized field - have their own specialized terms. And sometimes they don't match up. Naming is hard!
I'm not exactly sure where you're coming from, but in the world of 3D modeling topology refers to the way the geometry of the model is structured, including the placement and flow of vertices, edges, and faces.
bro, decorate your room
This place is a rental so I'm pretty limited to what I can do.
@@DECODEDVFX i'm just teasing you, great vids