Home Technical Talk

Sharp corner edges on a square detail over cylinder without distortion

Hii everyone,

As you can see, I would like to keep those corners sharp without any distortion over the mesh of the cylinder.

With the current edge flow that I did, a weird artifact is visible which is pulling the surface a little outward.

Any other method that I can use to completely eliminate such distortions?

Replies

  • sacboi
    Offline / Send Message
    sacboi insane polycounter

    "Any other method that I can use to completely eliminate such distortions?"

    one technique, is use the existing geometry as a support to control artifacting once the mesh is subdivided:

    https://polycount.com/discussion/comment/2762984/#Comment_2762984



    Edit: A topic among many discussed in the subd modeling thread - https://polycount.com/discussion/56014/how-the-f-do-i-model-this-reply-for-help-with-specific-shapes-post-attempt-before-asking/p1

  • poopipe
    Offline / Send Message
    poopipe veteran polycounter

    Is the thing you're trying to replicate a single watertight surface or made of multiple pieces?


    If it's the latter, you should probably make the model out of multiple pieces

  • Neox
    Offline / Send Message
    Neox high dynamic range

    quite an interesting tone to your post, considering you didnt solve the issue and still have quite some artifact on that corner

    but the thing is absolutely solvable with the geo you have in the left image, just turn some edges and leave some ngons :)


    while it is still not 100% gone, this way you can offset the issue sideways









    works better when softer but can also handle quite sharp angles before becoming too much of an issue










  • pior
    Offline / Send Message
    pior grand marshal polycounter

    Also FWIW, remember that one-click parametric bevels/creasing are also pretty good for ones sanity -


  • perprerp
    Offline / Send Message
    perprerp triangle

    Neox: "quite an interesting tone to your post, considering you didn't solve the issue and still have quite some artifacts on that corner"

    Neox, with respect, it is disheartening and discouraging to face criticism for my attempt to help someone. We all have different cultural backgrounds and neurological profiles and we all express ourselves differently. That is not for you to police.

    Perhaps you took issue with me saying that most artists get the principle wrong, Well, you got it wrong too:


  • Neox
    Offline / Send Message
    Neox high dynamic range

    Yep it was that "many do it wrong, here is how to do it right" i am not against you trying to help by any means. But your solution just as well had a massive shading issue.


    Its interesting how you have such a massive shading bug on my solution, i really can not reproduce it. As said its not perfect, but i only have a very minor artifact sideways of the corner, not as massively on it. What software are you using? I can imagine 2 things to happen. The bevel algorithm does something differently or the smoothing works differently.

  • perprerp
    Offline / Send Message
    perprerp triangle

    Neox: "Yep it was that 'many do it wrong, here is how to do it right'


    There are people here with autism, social anxiety, poor command of the english language, young age, or any number of diverse reasons for expressing themselves in a manner which you may not "approve" of, but perhaps we are here to discuss technical matters and it should remain strictly technical. Negativity breeds negativity. Nothing else. I wrote that most artists get the principle wrong not in an infantile attempt to stroke my own ego but to assure OP they were not doing someting unusually bad, and to highlight the importance of the principle. After more than 20 years of doing this I have some legit observations to make.


    Neox: "i am not against you trying to help by any means. But your solution just as well had a massive shading issue." / "Its interesting how you have such a massive shading bug on my solution, i really can not reproduce it."


    Look, both you and OP go against the principle I described and your meshes suffer as a result. I am not saying that to beef with you; it is just objectively, demonstrably the case. Your mesh would benefit from offsetting the extrusion the way @FrankPolygon and I did it. In this case it is not just an edges' angle which determines its shading, but its length too. It's not something anyone will argue with if they fully understand it.


    There are a *lot* of ways to handle the flow of geometry and how vertices connect. I have not touched on that at all. @FrankPolygon did.


    ilustration: showing different shading errors as the flow of a cylinder is interrupted by an additional edge. Providing that you intend to extrude the polygon to the right of the edge, I hope it is clear that you will want solution C over solution B.

    edit: forgot to say. I do not see any "massive" shading errors. They're well within what you get from top-end AAA assets.


  • Neox
    Offline / Send Message
    Neox high dynamic range

    i guess we can agree on the negativity argument. i think your point is valid, just the tone was off. schooling without essentially bringing the fix you wanted to show. i guess i shouldnt have chimed in on that tone... anyways what i'm trying to say

    this is not a crazy lot better than what OP had. a bit better yes. i guess is many cases the solution is valid. maybe one subdiv more would fade it into nothing. maybe on the final asset this will be hidden in roughness anyways.

    I offered another take on the issue, that usually works for me, i guess with pretty much the same amount of years in the indusrty we can agree that there are multiple ways of tackling an issue. offsetting it or solving it.

    With Frank probably offering the most clean solution here

  • parikshittiwari

    Hii, Sorry for the late reply.

    What you shared about using existing geometry as support edges, isn't that the exact thing that I did with my edges in the screenshot I sent? It definitely works but only to a certain extent.

    In fact, out of many options I tried this was the best in minimising the distortion and the distortion isn't even visible from the front. The problem arises only if change the view and look at the surface from a sharp angle. Also, if in case I needed an edge to be very sharp and tight then I would have to increase the geometry density of my cylinder very high. Wouldn't that make it extremely difficult to work on it for other details later on?

    I can also try and move the edges surrounding the extrusion but that destroys the curvature of the cylinder.

  • parikshittiwari

    I definitely always have the option to make it as a separate piece and place it there but I just wanted to know my options in a situation like this. I don't know why but I have always disliked placing different pieces together and not extruding or intruding them. I always try to make the details belonging to a mesh through the mesh itself.

  • parikshittiwari

    I understand your point. The screenshot I have posted was the best I could do and that reduced the distortion to a point where it wasn't visible if looked directly by the camera. It was only visible at a sharp angle.

    I just wanted to start a discussion over this and see what kind of ideas others could come up with. Was just curious....

  • parikshittiwari


    I might have to disagree on using tris at the corners. What you did on the left image is the thing I did in my screenshot and there are no distortions visible from the front unlike the one in your screenshot.

    As you can, my distortions are close to none compared to what is visible for you and it is more worse if I use tris instead of doing this. My main problem wasn't even distortions from the front. My problem was visible only when I look at it from sharp angles like the one in the screenshot I sent.

    I agree on what you said about keeping the transition geometry small, I understand the importance of it. I know smaller and denser geometry gives less distortion compared to bigger and less dense but in this case, what I did might also work and as I said, I need to eliminate distortions at sharp angles which will be visible in both the things you and I did.

  • parikshittiwari

    I think this was a mistake from my side where I couldn't describe exactly where the distortion is.

    The distortions are not visible at all from the front and when looked at directly through the camera. The edge flow that I have shown in the screenshot works extremely well in that regards. It is only visible when I look at the corner from sharp angles and that is the distortion I am trying to eliminate.

  • parikshittiwari

    They definitely are a godsend and creasing was always an option. I just avoid them right now because I am a student and I wanted to learn how I can work with the geometry in such cases. I like the support edges workflow more. I just think creasing might not help in cases where I want to export the model to a different program?

  • parikshittiwari

    Thanks for the incredibly in depth explanation you did. What you said is definitely the most cleanest way to go about modelling this detail. This also mostly eliminates the distortions visible from sharp angles.

    I will definitely take a look at the thread you mentioned.

  • sacboi
    Offline / Send Message
    sacboi insane polycounter

    "What you shared about using existing geometry as support edges, isn't that the exact thing that I did with my edges in the screenshot I sent?"

    No, not explicitly in terms of the example I'd linked.


    "It definitely works but only to a certain extent."

    Actually it doesn't and for my particular hard surface specialty I wouldn't even attempt too begin with when working via a subd workflow, anyhow subdivision algorithms are mathematically recursive, hence the reasoning why.

  • Neox
    Offline / Send Message
    Neox high dynamic range

    Your material is quite bright and matte, of course you will see less issues here

Sign In or Register to comment.