Home Technical Talk

How The F*#% Do I Model This? - Reply for help with specific shapes - (Post attempt before asking)

Replies

  • KebabEmperor
    Offline / Send Message
    KebabEmperor polycounter lvl 3

    A zoomed out image would help us see what are we looking at exactly

  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @guitarguy00 In the previous example: the base sphere has 12 rings with 24 segments and the subtracted sphere has 22 rings with 28 segments.

    Both quad sphere and UV sphere intersections tend to produce topology that requires some additional clean up. Minor differences in scale and position often prevent the perfect alignment of the segments on intersecting shapes. If the segments in the overlapping area are roughly equal then the mesh can usually be resolved to all quads, without needing the perfect alignment of each individual segment.

    Quad sphere topology is fairly consistent. Which does make it easier to match the segments of two intersecting spheres that are close to the same size. The down side is that the loop flow and total number of segments is quite arbitrary. This often produces inconsistent loop spacing around intersecting shapes that are larger or smaller than the sphere and also tends to produce triangles that can be difficult to resolve to quads.

    UV sphere topology produces distinct rings and line segments but terminates in triangles at the poles. The density of the rings and segments can be adjusted independently. Which does provide a lot more flexibility for segment matching than quad spheres but the concentration of triangles in the poles limits the usable section of the sphere.

    While there are workarounds like re-projecting [shrink wrapping] the subdivided mesh or generating quad spheres with unique segment counts, these processes aren't always quick or reliable. It's also possible to add or dissolve edge loops to resolve the mesh to all quads but this can produce smoothing artifacts like pinching and stretching. So, it's really about analyzing the shapes and deciding which type of sphere topology is best for that particular feature.

    The example below shows how both quad grid and UV sphere topology requires clean up and re-routing after the boolean operation. Mesh density determines the location and number of triangles left by the quad spheres. Ratio of rings to segments determines the position of the vertex pole left by the UV spheres. Adjusting the number of segments in the spheres will resolve triangles to quads and move the poles around.

    Maintaining relatively consistent segment spacing is an important part of creating curved surfaces that subdivide cleanly. While the geometry in overlapping quad spheres will generally match, it's the arbitrary loop flow that can cause smoothing issues. Especially when the size or position of the intersecting shapes causes the edge loops in the base sphere to bunch up or spread apart.

    Here's an example that shows how just changing the position of the overlapping quad spheres can produce a variety of smoothing artifacts.

    The quad sphere's grid topology is good for intersections that require multi-axis symmetry but the inflexibility of the loop flow can cause issues when the intersecting shapes are too large or too small for a particular area of the sphere. Which often means either manually moving the edges to make room for the intersecting shape or changing the density of the quad sphere by an arbitrary factor of two.

    In contrast to this, an intersecting UV sphere can be adjusted so the segments remain aligned. Which helps prevent smoothing artifacts caused by unintended shape deformation. It also allows both the larger sphere and the hemispherical pocket to resolve to all quads. The example below shows how UV spheres can be adjusted to maintain segment matching when the two perpendicular spheres are moved closer together.

    Here's another example of how the number of rings and segments in the overlapping UV spheres can be adjusted to control the mesh density in each feature and move the vertex pole around the inside of the hemispherical pocket. An additional advantage of the UV sphere's topology layout is that the rings produce straight lines that run perpendicular to the polar axis. Which makes it a lot easier to blend the sphere into rectangles and cylinders.

    It is possible to use a quad sphere as a cap for features on a UV sphere but the number of segments in the UV sphere needs to be adjusted to fit the adjacent segments in the quad sphere. The example below shows how the arbitrary number of segments in the quad sphere doesn't align well with unmatched segment counts. Even the mesh that's all quads has some pinching and stretching artifacts. This is because of the unintended shape deformation caused by re-routing the topology.

    A few of these smoothing issues could be reduced or possibly resolved by turning the triangles into n-gons. Whether or not that's acceptable depends on the technical requirements for the project. Adjusting the density of the quad sphere, by either (un)subdividing or manually adjusting the number and position of the edge loops, would be another option but could also produce unintended shape deformation. Which then requires additional work to either manually compensate for any errors in the shapes or re-project the mesh on to clean shapes.

    Adding or removing too much geometry will tend to produce unwanted triangles and can also generate surface deformations that will cause smoothing artifacts. Routing a significant number of triangles into a single vertex on a curved surface can also cause visible pinching. Manually deforming a simple mesh can work for some shapes but does tend to introduce a more organic feel and often produces shapes that are less accurate.

    The examples below show what these issues can look like. While it's generally a good idea to avoid these types of shape and topology issues, there are always edge cases. Subdivision modeling is about trade-offs. Just be sure to explore the alternatives and thoroughly evaluate the results with a quick block out. Before committing a significant amount of time and effort to a topology strategy that might not produce the desired results.

    When evaluating the accuracy and quality of a surface, be sure to use materials that will expose any subtle smoothing artifacts. Both low contrast and reflective materials that are overly sharp can hide different types of artifacts. This is why it's generally considered best practice to use materials with some contrast and either a soft reflection or wide highlight with enough roll-off to catch any minor imperfections in the surface. For high quality, reflective surfaces it also makes sense to use flow checkers to identify hidden surface quality issues.

    Here's an example that shows how a low contrast material hides a lot of the smoothing artifacts caused by shape and topology issues. The flow check material helps pinpoint the origin of waviness in the surface. The high contrast material also helps identify edge and surface artifacts that are lost in both the flow check and low contrast material.

    The comparison below shows how each type of topology layout is going to have it's own type of surface quality issues. Some of these artifacts are relatively minor and as long as they aren't noticeably visible to the players then they might be acceptable within certain project constraints. Not everything has to be perfect but part of making these technical tradeoffs is avoiding time sinks. Manually editing and re-working large sections of the mesh are time sinks.

    Most of the shapes and topology layouts from the middle to the right are quite marginal and should generally be avoided. While the manually generated shape on the right doesn't have any major smoothing artifacts it does have some significant accuracy and surface quality issues. Something that's less than optimal for hard surface projects that require crisp details.

    Overall shape is also really important because the forms play a significant role in what the topology needs to look like. One side of the recoil shield may need to be a quad sphere and the other may need to be a UV sphere with a segment count that matches the quad sphere.

    As designed, the 1851 is cap and ball. The loading gate is at the front. Below the barrel. There were some later conversion to metallic cartridges but they often used a modified version of the existing breech. Machine tooling of the era was pretty primitive and the shapes they produced were fairly basic. The prototypical hemispherical relief in the recoil shield is fairly small. Especially when comparing it to how much material needs to be removed for the rear loading gate. This type of conversion could be done by hand with a file or by machine with a basic boring or milling operation. It also left behind a bit of the original hemispherical pocket.

    The example below shows how the quad sphere's topology works well enough with the larger cylindrical cut out. It can also be mirrored vertically. Which would potentially reduce the amount of work required to clean up the topology around the shape intersection.

    As shown in the references, some of the original hemispherical relief cut tends to be left behind so that would rule out vertical mirroring but the spherical intersection is so small it's relatively easy to simplify the shape to get everything to line up.

    There's been lots of good discussion about different modeling methods and topology strategy. Just remember that the shapes are what define the topology. It's really easy to get bogged down in the technical aspects of 3D modeling but the important thing to focus on first is getting the shapes right. After that the topology can be adjusted to fit the shapes and there's so many different ways to approach the order of operations that the modeling tools are really down to personal preference.

    Recap: Block out the shapes, match the segments of intersecting shapes by adjusting the mesh density and rotating the topology to fit, clean up and re-direct the loop flow as required but try to avoid causing unintended shape deformation that can cause smoothing artifacts.

  • guitarguy00
    Offline / Send Message
    guitarguy00 polycounter lvl 6

    Thank you so much, alot of great information here. Would you ever recommend using Set-Flow/Edge-Flow to try and minimize any distortions in the sphere? Or does it usually fail because you are messing around with the even spacing of the native sphere?


    Also, i find it very strange that in 3DS Max, I can only adjust the amount of segments in a standard sphere, but not the rings.. Pretty disappointing.

  • Thanez
    Offline / Send Message
    Thanez interpolator

    @guitarguy00 You can alter the geometry, but it's a bit involved.

    So you make your Sphere/Geosphere demonchild, and that's when you notice it: The pinch.

    You try to mitigate it by chamfering that edge, but you have little control before the pinch gets even worse.


    Time to evaluate what the problem is. These edges come from the sphere and are "perfectly" flat on the X axis and of even length.


    These edges are remnants of the cube that made the geosphere, and are not flat on the X axis, nor of even length.


    Here are the three main culprits that end up as problems for us: First we get a curve that is supposed to give us a circular shape, but the edges that make up that curve are unequal in length. This will not make a circle when smoothed. What will make a circle is edges that are equal in length and have an equal angular variation between them.


    Second, these two edges both lead up to the intersection where a new edge will be "chamfered" in. For that chamfer to look even and nice, they need to be equally long.


    Third, and we can't really do much about them, are these polygons. They break up the edge flow and are the main culprit of the pinching. We can mitigate the effect by making them larger, which we will.


    It's morbin' modifyin' time. Go to your geosphere and select the first edge loop that is not a remnant of the cube you started out with, and while constrained by edge, scale to 0. This might seem unnecessary, but it no being flat on the X axis will negatively affect the result of next step.


    Next. Grab these pseudo-squares, cause we're about to turn them into pseudo-circles.

    Constrained by face, and "Use Pivot Point Center" as your reference, scale down until...

    ...this polygon almost turns into a triangle. Triangle-shaped polygons don't do well on spheres.


    I'm almost happy. These edges still have too much variation in length.


    Grab these four polygons, constrain by face, use "Selection Center" as reference point, and scale on YZ until those edges above look a tad better. Be mindful of the polygons you're now making into a pseudo-triangles at this point.


    Finally, add a spherify modifier to undo the hellish edits you did to this mesh and to make the geosphere actually spherical because they never were.

    Then do the steps to connect the sphere and geosphere.

    From left to right: Before any edits, chamfer severely limited. After making Pseudo-Circles. After scaling faces inward. Lastly a demonstration of how far we can push the chamfer now.


    "Hold up," you think, "did this guy just make the ultimate geosphere?"

    Nope, it only helps with transition into other shapes.

  • guitarguy00
  • KIimt1234

    Hi,

    I'm practicing retopology and I'm trying to figure out how to properly retopo this model. I've done it as best as I could figure out. It's mostly quads (except for a few N-gons I can't seem to get rid of), but I can't help but think that the edge flow might be 'messy'. Is there a way to improve it? I've attached the OBJ file and screenshots.

    Thanks


  • clark_tee

    I'm in the process of beginning to model an F-14 Tomcat and looking at the shape, the area in the pictures seems to be the biggest hurdle I see. I have also attached an image of a test I did. I am getting back into 3D modeling after a long hiatus and would love any ideas how to to model this area and avoid pinching and distortion without making that area of the model too dense. As you can see it needs to flow down to a point where it meets the rest of the aircraft boy and terminates smoothly on both ends. I am using Modo but the solution is not likely software specific.


  • HAWK12HT
    Offline / Send Message
    HAWK12HT polycounter lvl 12

    Hey @clark_tee

    looking at reference I am pretty sure wing and fuselage are not one connected body. Also for hard surface stuff continuous mesh is rarely a case in real life so its best to approach stuff like this in a way how its built (section by section)

  • okidoki
    Offline / Send Message
    okidoki polycounter lvl 2


    Thank you for this model ;-) tried some retopo myself.. have to triangles in (on half of the object.. resized and rotated it to my liking ~ 644 polys for a full object.. 802 was yours

    ).. Your big button on the back seems to have to need some more love.. it's disrupting the flow a bit too much. The general form does change a bit i my trial..


  • senwaaa

    How do I model Korean roof like that? I have tried lattice, triangle modeling and sphere editing to no avail and I just don't know how to do this.


  • wilson66
    Offline / Send Message
    wilson66 polycounter lvl 8

    I hacked something in Blender quickly, take a look at the attached file. This is roughly how I would do it. You probably use Blender already judging from the screenshot, but if not, you might be able to adapt the method to another app also.


  • senwaaa
  • solitudevibes
    Offline / Send Message
    solitudevibes polycounter lvl 2

    Hi guys i wanted to challenge myself and i finally picked a gun that has complex grip. Before that i was making guns with very simple grips like makarov pistol and 1911.


    So the question is how do i tackle this thing?


    I made the top part in fusion 360 without any problems but when the time came for the grip i could not progress any further. Ive tried subd method directly in fusion but it turned out bad. I tried making it in blender using subd workflow and my shading was broken on the low poly.



    it looks awfull because i wanted to figure out how do i optimize this thing for lowpoly. Before that my models were pretty simple and just removing the modifiers was enough to make the low poly. Clearly this workflow will not work with something like this. With round handle and square top part and this ergonomic indentation for a thumb. Whats the best workflow to make this thing? I need high poly and a low poly. Do i just need to apply all the modifiers and then manually clean it up?


    I use both blender and fusion 360 and i use zbrush only to make highpolies when importing from fusion.


    Please Help!


    Edit: also, it does not have to be subd ready model. I just need the shading on LP and HP to be good.

    my goal is to make it game ready

  • Kanni3d
    Offline / Send Message
    Kanni3d ngon master

    @solitudevibes

    Evenly spaced loops for predictable/smooth shading.

    Easy to edit/manipulate cage is key for these sorta organic parts. Basically less is more when starting forms out from large -> medium -> small. Don't have a lot of geometry as your base model.

  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @solitudevibes Block out all of the important forms and features that need to be visible in the base mesh. This will make it a lot easier to generate accurate high poly and low poly models that shade cleanly. Capturing all of the key details in the base mesh should solve the issue. Like Eric mentioned, an iterative block out process will work with most modeling workflows. Including boolean re-meshing, CAD, Subdivision, etc.

    https://polycount.com/discussion/comment/2731601/#Comment_2731601

    https://polycount.com/discussion/comment/2731727/#Comment_2731727

    Non-destructive or modifier based boolean re-meshing and subdivision workflows will work for semi-organic hard surface shapes. It just comes down to capturing the details in the base mesh and setting up the order of operations so the density of key features can be adjusted easily.

    https://polycount.com/discussion/comment/2768812/#Comment_2768812

    https://polycount.com/discussion/comment/2777056/#Comment_2777056

    https://polycount.com/discussion/comment/2776197/#Comment_2776197

    The same basic modifier based modeling approach can be used to generate base meshes for both re-meshing and subdivision workflows.

    https://polycount.com/discussion/comment/2768218/#Comment_2768218

    https://polycount.com/discussion/comment/2773644/#Comment_2773644

  • christrom

    Much appreciated @FrankPolygon for your insight. I do think that I should create a block out stage in future, as I can see it would help to see an overview of the flow and it would definitely help with planning. It is something I will look into further, as for the moment, I try to get one area perfect, and then move to another, which I realise is the enemy of efficiency. Its funny, but this is just like painting - it is arguably better to work on all areas of the painting simultaneously in a broad fashion, and then begin to refine in passes. I'm familiar with Andrew Hodgson's blog, and I will look into the other recommendations.

    I feel with time, block outs will become easier, as once the knowledge of Sub-D methods really gets emhe made more easily. I really appreciate your time

  • christrom

    Sorry, for some reason my thanks was not posted, so I've posted it now above, albeit very, very late. So again my apologies. I don't take advice and help for granted.

    I have a quick question regarding my process (should be v simple)

    Let's say I have the top and bottom vertex where I need them, and I want all the vertices going up the centre of the shape to follow these points. If I try to align the pivot to one of the faces, naturally it will throw the start and end point off when I scale the verts. At the moment I do the following:

    Move the pivot to the top vertex, and then using crtl and shift, aim it at the bottom vertex to create a custom pivot. Then I go into vertex mode, select all the vertices that need aligning, select custom pivot in the move menu, and try to scale on all axes until the line is straight.

    Like so. Seems like a long winded way of doing it though. Are there better options? I suppose I could snap, but that would possibly be more time consuming. Thanks very much.

  • Octavio
    Offline / Send Message
    Octavio node


    Hi I got this problem, can u tell me how can I fix that shading, the major problem⚠️ is that in the final bake u can see this same shading(I dont seethe need to post how it looks if it looks same than the 1st image)

    PD: I need that form of the cilinder (I can´t scale geo)

    hope u can help me :D cheers

  • Thanez
    Offline / Send Message
    Thanez interpolator

    @Octavio Please post a shaded view with wireframe on the entire area, as it's very hard to analyze any issues right now. Either that or an FBX.

  • Octavio
  • sprunghunt
    Offline / Send Message
    sprunghunt polycounter

    The lines you have are pointing to the polygon edges on the surface. If you go to the mesh properties and select "show hidden edges" you'll see there's a triangle edge there.

    You could try selecting all the vertexes and welding them with a low threshold. To make sure you don't have extra vertexes - this may help.

    But I'm guessing that the only way to fix this would be with extra subdivisions. Or a baked normal map from a higher resolution model would also fix the shading.

  • Octavio
    Offline / Send Message
    Octavio node

    @sprunghunt hello thx for the answer but If I make the bake from a higher model it would be apper in the final shading(with the bake applied) I also try to fix manually the normlas with the modifier Edit Normals and I don´t have a good result, the gral target is to preserve the low vertex count with a good final shading in that part ☺️ Chhers



  • soddak

    Hello, I'm modeling boot soles and the topology flow seems to be weird.

    Please help me......


  • KebabEmperor
  • okidoki
    Offline / Send Message
    okidoki polycounter lvl 2

    Maybe this isn't really needed ???


  • chopsuey
    Offline / Send Message
    chopsuey node

    Hi, I'm trying to train and I always encounter the same kind of problem, example with this simple model


    The best result I found is with this method but it's not perfect, I still see small deformations... but am I still getting close to a correct method?


  • Thanez
    Offline / Send Message
    Thanez interpolator


    @chopsuey Hey bud, this is the classic "cutting a hole in a cylinder" problem. I suggest you look through this thread for some learning, because this is just one method to do this.

    In short why you're getting stretchy no-no's is because you made an extrusion at the cylinder's established edge, and you are NOT allowed to alter the cylinder further without messing with it's roundness. What you should've done is keep that established cylinder edge as supporting geometry, and make your extrusion behind it.

    First off you need the correct amount of edges in your cylinder. For an extrusion that takes up about a fourth of the circumference, that number of edges is ~48, imo.

    Lookie, this is the area I want the extrusion in.

    I'll create some vertical edges that will house my real extrusion

    Then some horizontal ones

    Weld these corner verts

    Then make the extrusion

    Bing bang boom.

    In summary, cut between the lines, and use the cylinder's geometry as support geometry.


    Edit: Eh, fuck it, here's another method that preserves that girthy growth better.

    This one's easier to do, but at a cost. The highpoly has much more geometry in it, so may bake or render slower, take up more ram or storage space, depending on how you output this.

    Start with as few sides as possible, which is 12, because for a curve to be convincing, you need three edges to properly define it. I want the extrusion to be 1/4th of the cylinder. 3*4=12 sides to the cylinder. OKLETSGO


    Turbosmooth that mofo and until you've got enough geometry to support the extrusion, so just go ahead and make it using the polygons already selected. Oh and add this edge loop.

    Catchphrase.

  • wirrexx
    Offline / Send Message
    wirrexx ngon master

    adding to what @Thanez already wrote.


    you can also simply use a less dense mesh and move these edges to create even space


    you can also see that i work on 1/4 of the mesh. so less headache


  • Thanez
    Offline / Send Message
    Thanez interpolator

    @Octavio Sorry I took a while, but here's my two cents.

    I checked out your model and I can't find any fault with it, it's working as intended, and you didn't do anything wrong. The shading you see happens because when you scaled that smol oval into the big oval, you also squished it.

    When you squish a circle like that, you make the edges of the polygons non-planar, revealing the fact that all quads are made of triangles. Turning on facets shading mode illustrates the issue perfectly. Here on the right, I stole your oval shape and remade the big part but without squishing it.


    Here's looking down the barrel of those polygons, showing how yours aren't flat, while mine are.


    Luckily this doesn't affect the model's ability to be used as a cage for a highpoly.


    If you look real closely you can still see the triangles in your faces here, but the effect of it is divided by the amount of polygons that now make up the transitional cone.


    When i turn off facets shading, it all disappears and smooths nicely. There's no way to tell yours apart from mine except that yours is flatter.

    This means you can bake that smoothness into a normal map.

    Edit: There's also a different factor, though it doesn't have as a big effect. The best way I can describe it is that we both have compound curves made up rectangles, and that sort of wobbles the smoothing a bit.


    The only way to mitigate this is to add a bunch of edge loops until the compound curve is made up by as square polygons as possible.

    Both are still good for highpoly, and should be baked down, as the main issue here is that this is a complex shape that's gonna need a lot of geometry to smooth nicely. I hope this answers any questions you had, if not, do reply.

  • chopsuey
    Offline / Send Message
    chopsuey node

    @Thanez thank you for taking the time to answer me.

    For your first example, I had already made similar tests, the problem is that by adding this edge loop we modify the shape of the object.

    The second example is interesting and gives a better result because it doesn't deform anything. I'm on blender maybe it's different on 3ds max but to do that I have to apply my subdivision modifier which is the equivalent of turbosmooth if I understand correctly and I don't like this idea very much if I can do otherwise.

    @wirrexx Thanks, I like this technique, it's a good compromise that gives a very good result

  • Octavio
    Offline / Send Message
    Octavio node

    @Thanez

    Hey Thank you so much I really appreciate your answer, the last question that I have. How did u fix the non planar problem, can u share, wiith details please. I´ve tried to make the faces planar one by one but in my final result the "circular border" that I get looks so bad.


  • Thanez
    Offline / Send Message
    Thanez interpolator

    @Octavio I didn't. I made the same shape you did, except I didn't squish the large oval, meaning my cone didn't get squished on one side, meaning my cone had planar faces while yours didn't. I did that only to show you that both of them would meshsmooth nicely, and that you didn't do anything wrong. Also, the shape you're asking for is impossible. You can't squish one side of a cone and ask for it's faces to be planar. You can keep hitting the make faces planar button but all you're doing is trying to force a squished cone to not be squished, and you'll ruin both ovals and the cone.

    I tried to explain in my previous post that you didn't do anything wrong. Non-planar faces in compound curves happen all the time. They do bring some smoothing issues to the table, but that's in part why we bake normal maps. As long as your highpoly looks nice, and your lowpoly is ready for baking, it'll look nice ingame.

    The smoothing algorithm that smooths the reflections across faces has it's limitations and boi you have found them. Your model is perfect, it just needs a bit more geometry. Slap some supporting edge loops on that bad boi, meshsmooth it and bake the normals.



    However, if you really want to be a stricler about the planarity of faces, the only solution I know to something like this is to boolean the pieces. Remake the cone so the faces are planar, and remake the big squished oval as a separate piece.


    Then boolean out a transition between the parts. This will maintain the planarity (if that's even a word) and relocate the angular difference to the intersection between the parts instead of the faces of the cone.


    Add some edge loops to conserve the geometry of the cone and squishy, and weld the verts in the intersection.


    And would you look at that, it smooths nicely, even with those n-gons in the intersection.


    And I would leave it there. Now it looks like a helicopter rotor wing thingy with a nice CNC-looking intersection.




    How. ever. I feel like a person that cares this much about the planarity™ of faces won't be happy about N-gons in the intersection even though it's Good Enough™

    So we go back to step one but we're gonna need more geometry because with this method, the curve of the squished oval will be robbed of geometry at it's sharpest point. We'll start with a 32 sided cylinder instead of 16.


    We redo the boolean

    And delete the faces of the squished oval.

    The squished oval was only useful to give the cone's vertexes their position on the x axis that they needed in order to conform to the faces of the squished oval.

    Now go to vertex selection and select any vertex that's only connected to 2 edges. These vertexes belong to the squished oval. They are traitors and need to be dealt with. Delete them swiftly.


    Now, simply extend the border edges on the x axis to remake the squished oval in the cone's image.


    Add some edge løøps


    All the faces are planar


    It smooths nicely

    Glam shot


    This, like all things, was not free. We sacrificed the consistency of geometric density on the squished oval; We took edges away from the sharp point and gave it to the top point.

    This is why I said to go with 32 sides instead of 16. At 16 this would be a knife, not a squished oval.


    But I digress. Your first attempt was Good Enough™. My first and second attempts were Good Enough™, and this last one is Good Enough™. In the end they all come at a cost.

    There are many roads to rome. Travel them all and you will gain knowledge. Try to find the perfect road, and you will only find insanity. Poemtry.

  • acarmona88

    Hi Guys!

    I need some help with a little issue I am having with a model, I know that maybe it is easy, but I can't find the way to solve it or to get a better polygon flow.

    Baisicaly I am modeling something similar of a screw but I am ending having this little peaks that I can't fix, if someone could give me some tips it would be awesome, and I would be very very greateful!


    Thank you for your time!!

  • hanabirano
    Offline / Send Message
    hanabirano polycounter lvl 4

    @acarmona88 this shape is basically a cylinder with rectangle intersections, here is some info on how to do these shapes: https://www.artstation.com/blogs/frankpolygon/o7Pg/sub-d-modeling-cylinder-and-rectangle-intersections-fillets

    the only difference is the rectangle intersection pushes inwards

  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @acarmona88 Welcome to Polycount. Consider checking out the forum information and introduction thread.

    Sharpening unsupported corners on curved surfaces often produces smoothing artifacts that can be resolved with a few different modeling and topology layout strategies. Deciding which is approach to use really just comes down to figuring out how accurate the surface needs to be. Something that's often determined by how close the model is to the viewer.

    Like @hanabirano suggested: look for examples with similar shapes and try applying those same strategies. This thread is a great place to start and there's often several examples of how other artists have solved similar problems in different ways. Below are a few links to some more in-depth write-ups about square cut outs in curved surfaces, managing how loops cross surface transitions and placing details on or between the edges of a curved surface.


    Topology strategies for square cutouts in solid and hollow cylinders:

    https://polycount.com/discussion/comment/2757713/#Comment_2757713

    Managing support loop paths across curved shape intersections:

    https://polycount.com/discussion/comment/2769713/#Comment_2769713

    Placing details on or between the edges of curved surfaces:

    https://polycount.com/discussion/comment/2772925/#Comment_2772925


    With subdivision models, it's generally considered best practice to use the existing edges of curved surfaces as part of the loop path for the support loops that sharpen the corners. Something that often requires adjusting both the loop flow and mesh density. Keep things relatively simple when blocking out the shapes and use the geometry of the primary forms to guide the loop flow.

    Here's just one example of what the basic modeling process could look like: Create a basic spiral shape then connect the adjacent edges to define the vertical section connected to the base. Fill in the empty space and use an inset operation to create the basic loop path around the outside of the shape. Delete the left over faces in the cutouts. Generate the rest of the model with modifiers like solidify for depth, bevel for the support loops and subdivision for smoothing.

    This topology layout uses the same pair of loops to control both the sharpness of the corners and the width of the edge highlights. Horizontal loops can be adjusted up or down and internal loops can be adjusted in or out.

    Whether or not this level of surface accuracy and sharpness is acceptable depends on how closely the object will be viewed. Subdivision modeling is inherently imperfect. So, there's almost always tradeoffs. Sharpness and surface quality can be increased by starting with a higher number of segments in the spiral but this can also decrease the overall editability of the mesh.

    Sharpness can also be increased by adding another pair of support loops around the existing loop path but this can decrease the overall accuracy of the surface.

    The previous mesh is passable for most situations but there can be some subtle deformation artifacts that may appear when the model is viewed or lit from glancing angles. These types of artifacts are often cause by support loops that disrupt the segment spacing and produce subtle undulations in the surface. Sometimes it is possible to minimize the visibility of these artifacts by manually adjusting positions of the corner vertexes but this can degrade the overall accuracy of the surface.

    Whether or not this type of tradeoff is acceptable really just comes down to whether or not the artifacts are visible. Low gloss or matte materials with lots of high frequency surface textures will generally cover these shallow artifacts. Offsetting the position of the cut outs or adjusting the segment density of the starting mesh can also improve the surface quality. Without having a negative impact on editability.

    The position and width of the support loop path can also be adjusted to fit different size cutouts. Sometimes it makes sense to have a face or an edge between two perpendicular loop paths. Use whichever approach is easiest to work with and still produces clean results. Same for deciding how many segments to use.

    Some types of projects do require quad grid geometry but if there isn't a specific technical constraint requiring it then it's often fine to use a few triangles or n-gons that are well supported and aren't causing visible smoothing errors. It just comes down to those tradeoffs between accuracy and editability.

    Recap: Use the primary forms to route the loop flow. Use an appropriate number of segments to support the shapes at the desired level of surface quality.

  • acarmona88

    Hello again guys!

    @hanabirano and @FrankPolygon thank you a lot for your aswers!!

    I really helped me a lot! I spend a lot of time searching on the internet, and checked a lot of posts in this same thread but I coudn't find a solution so I decided to post for some help.

    I always knew that I shoud practice more my modeling skills, precisely for this type of "hard" modeling objects. And @FrankPolygon I must say that your answer was impressive man! Thanks for your time!

  • Octavio
    Offline / Send Message
    Octavio node

    @Thanez

    So I´ve tried to make every single step that you told me and I can´t figure it out how u made the boolean :s. I´ve had apply the intersection with ProBOOLEAN

    First I build the geo

    But I have this result:

    And what I want is what u had show me,


    I know that I just can select the loop and extrude. Support loops and thats it.

    Cheers. and sorry if I already bother you

  • Thanez
    Offline / Send Message
    Thanez interpolator

    @Octavio You could never bother me <3

    It was unclear in your post if you still had issues. There's a critical step of removing some verts that belonged to the original squished oval. When you're at this stage:

    Add an edit poly modifier to the boolean, (a) go to vertex manipulation mode, (b) go to Selection in your ribbon, (c) by numeric, (d) using 2 as input, then hit backspace. That'll remove the verts but keep the edges of the cone.


  • Octavio
    Offline / Send Message
    Octavio node

    @Thanez Super Glad to hear it!!!

    Oh! thanks but what I mean is that I can not make this step

    You have all this shape

    This is my setup and my result:

    Result:

    A: Settings Can u share your settings in order to have your result? B:Result


    This is the part that I missed

    Cheers.

  • Thanez
    Offline / Send Message
    Thanez interpolator

    @Octavio Ah, that's because I capped all the open areas before doing the boolean purely out of habit. The end result is the same, you just don't have to delete the bits I did.

    I did notice that I got some artifacts on my result. The problem was that my squished oval that I booleaned with was too low poly, making some of the verts of the cone out of place. I redid the whole thing with a 64 sided squished oval. The cone is 32 still.


    You can have the model if you wanna peruse it but you'll have to wait a day because I, an intellectual, forgot to pay my hosting bills, so my webzone is down. I'll edit this post with some sort of squished oval cone transition or something.max in case you're interested

  • HAWK12HT
    Offline / Send Message
    HAWK12HT polycounter lvl 12

    Hey all,

    Long ago someone here shared the cheat sheet to use for number of sides but I cant find it on older pages, so like in example above

    shape on left has 20, 16 and 12 sides for large medium and small details, changing the radius to align the segments and running boolean required minimum cleanup and subdivides ok. (Got this from previous posts)

    On right I tried with 24, 20 and 16 thinking 4 being a magic number and it all fall apart for the smallest shape. I can make it work by lowering its sides and changing radius however it will not always be ideal to scale the parts if they dont match reference. I want to understand the maths behind choosing sides for intersecting cylindrical details. Thank you.


    EDIT: I explored further and made it work, there are slight inaccuracies but not enough to impact normal map bake. Still need to figure out why certain number of sides work.


  • KebabEmperor
  • HAWK12HT
    Offline / Send Message
    HAWK12HT polycounter lvl 12

    Thank you, I follow Frank's guide religiously however I am looking for that cheat sheet / trying to understand the number logic (20,16,12 sides) as shown in my post. It was shared very long ago and the images are not showing on older pages here.

  • KebabEmperor
    Offline / Send Message
    KebabEmperor polycounter lvl 3

    Not sure what that is but it is usually about size of the cylinder (half the size, half the segment count) + experimenting. Not sure if that's maya or max but in blender we use screw modifier to be able to change segment counts to match other cylinders so we can try & fail easily.

  • Dinacatrom

    Hey guys, im trying to get a clean typology here but i've tried a bunch of different approaches but im kinda stuck.

    im trying to model these air vent things on a locker but i cant really figure out what the best approach is here.

    ive watched a couple tutorials but they all do it differently and non really work for me, so it would be appreciated if somebody could lead me in the right direction :D


    thanks in andvance!

  • ZacD
    Offline / Send Message
    ZacD ngon master

    You just need a flat evenly spaced loop around the vent before connecting it back to a flat grid.

  • Dinacatrom

    i'm sorry but im not sure if i understand... i tried but it turned out like this:


  • okidoki
    Offline / Send Message
    okidoki polycounter lvl 2

    Maybe like so (all quads):


  • wirrexx
    Offline / Send Message
    wirrexx ngon master

    nothing wrong with that, you just need to make sure that you have the supporting loops at the right places and thikness would help too



  • Raphael_Bouch

    Hey folks, I am trying to model this Viper MK VII from Battlestar Galactica but I can't bring myself on how could I combine all these booleans together

    I tried to make the topo line up with loops and stuff but it still turns out rather messy after clean up.

    The rear also has to be empty to put the engines inside and I don't know if doing it the boolean way is wise or should have I made it all one piece ?

    Thanks in advance for any help !

Sign In or Register to comment.