Home Technical Talk

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

Replies

  • naman
    Offline / Send Message
    naman node

    How to model this ?

    I am working on this gun part but after subdiving it's resulting in a strech in the model. Can anyone tell me how to fix this even after subdividing.

  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @naman Resolving this type of smoothing artifact is a fairly common question and there's a number of different ways to approach the order of operations and topology routing. This thread has lots of great examples provided by the community so it's generally worth the time to have a quick look for how other artists have solved similar shape intersections.

    A few of pages back @wirrexx shows how to blend similar intersections into spheroid shapes using n-gons and a couple of pages back there's a write-up that covers several different solutions for merging edge loops from shape transitions into the sides of cylinders.

    Most approaches to resolving this type of smoothing artifacts can be boiled down into either using the existing geometry as a support loop by offsetting the intersecting shape or terminating the perpendicular edge loops in a triangular quad or n-gon.

    Which approach makes the most sense will depend on whether or not the shape is constrained by adjacent geometry, how sharp the edges need to be and whether the transition into the cylinder needs to be exactly at the mid point. The following example highlights the basic concept and both approaches can be resolved to all quads if required.


  • Deqa
    Offline / Send Message
    Deqa polycounter lvl 8

    Having trouble with the chamfers on this truck cabin. I start from a cube, chamfer or bevel all 4 edges of the cube, and then try to inset for the window, inset for the driver side window... but then I realize, it's really hard to control my bevel loops, subdivision, and get it to look like the truck cabin from the reference. It ends up just looking like my attempts below.

    Especially this flat transition from the cabin's chamfer to the window. Ultimately, it's a mess of too many edge loops, and too many chamfers / bevels to create this cabin's curved edges and flat transitions!

    Any help would be appreciated to try to get this closer or what order of operations I should follow. Chamfering a cube first and then trying to inset for the windshield or adding control loops for flat transitions doesn't seem to work.


  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @Deqa Automotive modeling is it's own specialist discipline but most of the basic principles of subdivision modeling are still relevant. @sacboi has provided some helpful guidance and links to some great write-ups about car modeling in a recent discussion.

    It's often helpful to block out the larger shapes first then confirm that the mesh subdivides cleanly before adding secondary details, like cutouts for the doors and windows. Below is an example of what the modeling process could look like.

    Break down the shape of the cab into individual planes then round over the transitions. Keep the geometry relatively consistent yet simple. Edge loops can be cut in and dissolved as required but flat surfaces should be kept co-planar and curved surfaces should have uniform segment spacing whenever possible.

    Since these types of trucks tend to have a lot of flat surfaces, the same sort of block out process can be used to create a variety of different cab shapes. The important thing is to focus on creating accurate surfaces that are co-planar and consistent transitions that are fairly smooth. Keeping things relatively simple during the block out will also make it easier to solve shape and topology flow problems.

  • naman
    Offline / Send Message
    naman node

    Someone help me on this. How to model this without strech.


  • ZacD
    Online / Send Message
    ZacD ngon master

    I'd model separate pieces as separate models.

  • laeion

    @FrankPolygon

    Thanks for take your time to answer! I think you gave more answers than questions I had :) ! I had an idea for your method (manual adjusting mesh) but my first tought was, that nobody do this way, .. but then I was wrong! I hope that I will find a little time, to try these modelling methods! I follow your threads, but honestly never heard about soft hard surface modells, I think I found something new to learn!

    @snakedogman

    Well, I'm not working in the game industry, so modelling is a hobby for me at the moment. I don't make my modells to commercials or games, therefore my goals are different. I more like to create stylistic renders/animations instead of photorealistic. So I targeting more something between high poly and low poly. (mid poly?) I don't know this is a good approach, but since this is a hobby project for me I used to challenging myself to modelling something with minimal polycount but over lowpoly.

    If I created some useful modells I will post them here for some critique!

    By the way the style what really catch me is this (Arseniy Chebynkins work:

    https://www.artstation.com/artwork/8eG8Dw

    https://www.artstation.com/artwork/o2EoB

  • solitudevibes
    Offline / Send Message
    solitudevibes polycounter lvl 2

    Please, someone help. I just cant wrap my head around on how do i model this?

    Last image is what im aiming for


  • solitudevibes
    Offline / Send Message
    solitudevibes polycounter lvl 2

    Hi guys, ive reached a dead end and i cannot progress any further.


    IIts the bottom part of the handguard that gives me so much trouble. The curve at the bottom and the inset that ends with 2 edges connecting into a single vert. I tried extruding and using booleans but it just does not work. Please, tell me how do i even start making this absolute nightmare of a model

  • hututuzhang

    Hi guys,How to make it round?


  • ZacD
    Online / Send Message
    ZacD ngon master

    Use more edge loops, extend the loops down and through the hole instead of looping around the corners, that'll pull out the corners like you're seeing.

  • hututuzhang
  • bittermelon

    Hello everyone. This is my first time trying to model something but I was wondering how I would be able to get a sharper edge on this? I'm trying to create a Pokeball. Thank you


  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @solitudevibes There's a couple of different ways to approach modeling a lighter hood. A direct approach would be to block out the major forms with a simple quad grid then subdivide to create the necessary support geometry for the holes. @wirrexx explains this modeling process, with a great visual example, in another thread. Though the shape of that hood is slightly different, the same basic principles can still be applied here.

    It's also possible to model the shape using floaters or boolean operations. Which approach makes the most sense really depends on the technical requirements for the final model. There's a few recent examples of how to add details to curved surfaces using these modeling strategies. So, it's probably worth taking a look at some of the previous pages in this thread and finding some write-ups that show how to add circular cut outs to curved shapes.

    For the hand guard: When modeling objects with complex shape intersections, it can be helpful to start the block out by analyzing the references and color coding the major forms and important shape transitions. Finding and studying reference material, like drawings, images, videos, etc., is an important part of the modeling process. Gather enough reference material to develop a working understanding of the relationship between the shapes that make up the key features and the shape transitions between the major surface planes.

    Keep the initial block out relatively simple. Focus on creating the larger forms first then start adding smaller details. Maintain co-planar geometry for all of the individual surface planes that were identified previously. Continue working through all of the forms in the references. Try to resolve most of the major topology flow issues by matching the segments of intersecting shapes. Additional edge loops and final support loops can be added once the block out is complete.

    Below is an example of what this process could look like when using booleans to create the primary features and bevel / chamfer operations to generate the curved shape transitions. The top cover and a few shapes on the inside have been omitted for simplicity.

    Analyzing the reference images, identifying the shapes that make up key surface features then constructing the surface planes and generating consistent transitions between the shapes is a large part of hard surface modeling. As long as the geometry that defines the shapes remains relatively uniform the shapes themselves should define most of the loop flow. Which is why time spent gathering references and working through multiple iterations of the block out phase is usually paid back later in the modeling process.

    While some shapes are relatively obvious, it can be helpful to get a second set of eyes on the references, shape analysis and block outs. Posting shaded and wireframe images of previous modeling attempts makes it easier to provide focused feedback. Which is an important part of working through difficult shapes that aren't turning out as expected. Though often much slower, it's also helpful to look back at previous attempts and break down the process to find what worked and what didn't. This sort of self reflection isn't always fun or easy to do but it's a significant part of growing as an artist.

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

    @hututuzhang There's a write-up on the previous page that covers a similar shape. This thread has a lot of great examples of different approaches to modeling. While some of the examples won't match any given question exactly, most of the basic modeling principles are the same and the order of operations can usually be modified to fit a specific 3D DCC. So, it's generally worth the time to take a look back and try to find a discussion about similar shapes or topology problems.

    @bittermelon A few recent discussions in this thread have touched on cutting shapes into spheres and hemispheres but one of the simpler answers is to use a pair of perpendicular support loops that run around the perimeter of the shape and cross near the corners. Moving these support loops closer to the edge of the cut out will tend to sharpen it. Below is a basic example of what this could look like. Other, application specific, approaches like creases could also be an option but may not work well for certain types of modeling workflows.


  • hututuzhang

    I saw it, thanks.I imitate your wiring,The problem has lessened, but it remains.

    I would like the effect to look like this,Or maybe this is an effect that cannot be done?


  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @hututuzhang The distortion is caused by reducing the loops instead of carrying them across the shape.

    While it is generally possible to manually compensate for this type of shape distortion, some broader context is also required. Catmull-Clark subdivision smooths by averaging the existing vertices. This recursive smoothing tends to be visually appealing but can also reduce the accuracy of the shape once it's been smoothed.

    If a high level of accuracy is required then the starting geometry of a subdivision model will generally need to be quite dense. Which can be difficult to work with when using poly modeling tools. This is why it's generally easier and more efficient to work with parametric modeling tools, like NURBs or other surfacing tools in CAD applications, for these kinds of tasks.

    Most game and VFX models don't require this high level of surface accuracy and the relatively minor accuracy issues inherent to subdivision modeling are generally acceptable. Softer shape transitions aren't necessarily a bad thing either. Sharp corners that are exposed tend to be knocked down or chipped off during general use. Since the shape distortion in this example is constrained to a very small section at the top of the bore and is relatively subtle, it's actually a fair representation of a visually realistic surface.

    The example below shows how this type of subtle deformation is generally constrained to the area between the two support loops. An easy way to resolve this unintended shape deformation is to move or scale the corner vertex outwards. Which will compensate for the smoothing stress that's pulling it backwards. Keep as many of the existing surfaces co-planar as possible when making these kinds of manual adjustments. This will help prevent creating any additional smoothing artifacts.

    Most models won't be viewed from strictly isometric viewpoints either. Which is why it's important to consider the player's average view distance when making decisions about the width of support loops and whether or not minor smoothing inaccuracies are worth resolving. The flip side to all of this is that once the edge sharpness drops below a certain size, relative to the overall scale of the model, subdivision modeling starts to make less technical sense.

    The example below shows how narrower support loops concentrate the additional geometry in a smaller area. Requiring more work to resolve the shape inconsistencies caused by disrupting the segment spacing of the intersecting cylinder and potentially introducing other types of shading artifacts.

    If this level of deformation is unacceptable, especially with this tight of an edge highlight, it may make sense to increase the amount of geometry in the base shapes or if surface accuracy is more important than visual readability then it may make sense to look at alternate modeling workflows that don't smooth the shapes by averaging the existing vertices.


  • solitudevibes
    Offline / Send Message
    solitudevibes polycounter lvl 2

    @FrankPolygon amazing breakdown! thank you, Frank

  • sacboi
    Offline / Send Message
    sacboi high dynamic range

    Cheers Frank, nice of you to mention other contributors too this thread.

    Edit:

    Also imo, that weapon foregrip looks very juicy indeed 👍️

  • hututuzhang
  • Raphael_Bouch

    How the f* do I combine these ?

    Hi ;) ! I am modeling a radial engine cylinder and I have some real issue combining these "heat fins" into the main cylinder corpus in a subd friendly way. As you can see the fins are pretty dense and their position seems incoherent with the cylinder body although I they are placed exactly as they should be based on blueprints.

    Here is the .blend in question if anyone wanna play with it, thanks in advance !!!

  • Kanni3d
    Offline / Send Message
    Kanni3d ngon master

    You've kind of answered it yourself - in order to combine such fine/thin details onto a large, not very dense mesh - your cylinder needs to have a lot more segments to support that sort of intersection, which in return gives it some supporting geometry without pinching.

    I'd advise simplifying, and thickening the fins up (for more readability, and so that you don't have to go too dense with your cylinder)

  • hututuzhang

    Hey guys, I want to ask how to model this?


  • naman
    Offline / Send Message
    naman node

    Hello, I want to create this without any hard line. Can someone help me with this. It should be one object. I have tried multiple ways but not working.


  • wirrexx
    Offline / Send Message
    wirrexx ngon master

    do you have a better reference. If i understand it correctly, you want the cube part that goes into the cylinder to be softer? just even out some of the edges that are close to the area more.



  • naman
  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @naman Three fundamental concepts of subdivision modeling are: use the existing geometry as support for shape transitions, use a reasonable amount of geometry to generate the shapes, and match the number of segments in adjacent shapes to maintain uniformity. It's also important to either route support loops across the existing edges or simplify the loops by terminating them in the space between perpendicular loop intersections.

    This cylindrical cut out is similar to the cylindrical box intersection and the topology routing solutions are essentially identical to previous examples.

    All three of the fundamental subdivision modeling strategies are still relevant when the part changes orientation or the cut out changes position.

    Posting a clear example of the topology layouts attempted and the resulting flow or smoothing issues will make it much easier to provide relevant feedback. Not only is this an important part of self directed learning but it's also good practice to be respectful of other artists' time. A clear description of of the issue(s) and the desired outcome also makes it a lot easier to understand the constraints and goals for the model.

    There's a lot of great examples of different shapes and topology flow solutions in this thread. While it can be time consuming to search page by page, it's also possible to search for shapes using Google's image search. Just remember to refine the web search using site:polycount.com how do you model this and the name of the object or a description of the shapes.

  • sacboi
    Offline / Send Message
    sacboi high dynamic range

    @naman said:

    Hello, I want to create this without any hard line. Can someone help me with this. It should be one object. I have tried multiple ways but not working.

    No not this one. Help me on this.


    Hey...apply some manners coupled with common courtesy when asking for guidance since you hadn't even bothered posting an attempt to begin with plus I'll also strongly recommend binning that obvious sense of entitlement! - too

    Bluntly put, you're actually quite lucky receiving feedback at all especially from experienced people, who've taken the time and effort to offer help, aside from this specific instance but for many other queries over the preceding years, as well.

    So I really think an apology should be forthcoming, don't you?!

  • naman
    Offline / Send Message
    naman node

    Thanks @FrankPolygon for this. Last time you helped with that cylindrical extruded shape, that technique is awesome.

  • naman
    Offline / Send Message
    naman node

    I didn't recall any moment where I misbehave with someone. Have some manners. Don't try to get attention by doing this. If you have some problem with me message me personally.

  • naman
    Offline / Send Message
    naman node

    I agree with all this. But I didn't recall any moment where I misbehave with anyone in this group. Don't try to get attention by doing this. If you have any problem with me message me personally.

  • hututuzhang

    Hey guys,why is there a shadow here?

    😭


  • naman
    Offline / Send Message
    naman node

    Try selecting the object.

    Go to object data properties

    Normals - turn off auto smooth

    Or

    Select whole object and shade flat.

    Hope this helps.

  • hututuzhang

    Thank you very much!😁Normals - turn on auto smooth,It looks normal now.

  • SkinnyM

    Hello everyone!

    I faced with may be a trivial case, but solution is not obvious to me.


    I see only 2 ways to deal with such an angle.

    First:

    But here I have the quad which has a shape of triangle. Is it ok?

    Second:

    And here I have a star that shouldn't be there.

    Is there any other method to solve this kind of angle?

  • sprunghunt
    Offline / Send Message
    sprunghunt polycounter

    I would normally do the second version but extend the edge loops

    like this:

    at the very least I'd extend out a single edge loop like this:


  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @SkinnyM Both topology layouts are generally acceptable for most hard surface models.

    While the quad grid corner is a more conventional choice, the triangular quad corner tends to produce more appealing results for this type of shape transition. Something that can be verified by previewing the subdivision and looking at the mesh from multiple angles. Comparing and validating different topology layouts also provides an opportunity to learn which loop routing strategies produce the desired results and when it makes sense to use them.

    Although most flat surfaces are relatively tolerant of less than ideal topology layouts, as @sprunghunt mentioned, carrying some of these loops across the mesh does tend to produce more consistent smoothing behavior. Which is why segment matching and controlling shape deformation, that's often caused by abrupt changes in the loop routing, is an important part of preventing smoothing artifacts on high quality surfaces and curved shapes. So, although it isn't always essential for certain types of hard surface shapes, it is generally considered best practice to have relatively consistent loop paths. Even on flat surfaces.

    It is possible to simplify the quad grid layout by reducing some of the edge loops into n-gons or triangles but since the triangular quad layout tends to be more efficient this type of topology strategy is often limited to specific edge cases. Context is also important. Something that's passable in one situation may not make sense in another. Topology requirements for game art tends to be slightly more relaxed when compared to other fields like visualization, VFX, etc.

    Recap: Test different topology layouts and compare the results. If the mesh subdivides cleanly and fits the technical requirements for the project then it's generally going to be passable. When it comes to hard surface subdivision modeling, focus on developing the proportions and shapes first then move on to work through the base topology flow before adding the support loops. This will make it much easier to solve loop flow issues and optimize the mesh.

  • SkinnyM

    @FrankPolygon  great big thank you for your another complete answer! And for the link to another comment, it helps a lot.


    The edgle loops are not extended on my screenshots just because the question was only about that corner :)

  • sprunghunt
    Offline / Send Message
    sprunghunt polycounter

    If you're using a typical edgeloop insertion tool to create those loops you'd need to go back and delete edges to make the second example I gave. It's quicker to just leave the extra edgeloops in there.

  • naman
    Offline / Send Message
    naman node

    Hello guys,

    I am working on this part. But even after trying different ways I am not able to make it right. Please help me on this part.

    I want to remove these strech lines from my mesh.

    I want to create this part in one mesh.

  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @naman Looks like the subdivision artifacts are caused by a combination of issues. The curvature at the bottom of the guard is deformed by the intersecting shape and the extraneous loops that run across the intersection are pinching in the corners. Similar to the previous examples: both of these issues can be resolved by using the existing geometry in the curve as support and terminating the extraneous support loops with a triangular quad, n-gon or triangles.

    Another option would be to increase the number of segments in the curved section. Which would provide a path for each of the support loops on the intersecting shape but would also make the base mesh too dense to edit efficiently. These results aren't much better than the previous topology strategy. So, all of the added geometry doesn't really make sense on something so small.

    Most smoothing artifacts on curved surfaces can be resolved with the same basic subdivision modeling strategies: line up the segments in the adjacent shapes, use the existing geometry of the curve as support loops for the base of the shape intersection, and either join the support loops from the shape intersection into existing segments on the curve or terminate unneeded support loops between the edges at the base of the shape intersection.

  • naman
    Offline / Send Message
    naman node

    Thanks for the help. Applied this method but triangular quad end is not working as expected. Can you check on this.

    Here the file link - https://drive.google.com/file/d/1YqKVCPLbCNOO20SmGHPhDLleUSdpEFsa/view?usp=share_link

    Or this much is ok.

  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @naman The support loops aren't wide enough to allow for a smooth transition where the triangular quads were used to reduce the perpendicular loops. Whether or not the smoothing artifact is acceptable really depends on how accurate the model needs to be. Which is usually based on the intended use for the model and how closely this part of it will be viewed.

    Most of the edges in the subdivision model are extremely sharp. Which can make some of the shape transitions difficult see from a distance and will likely cause baking artifacts. If the model is going to be used in a high poly to low poly workflow then it probably makes sense to use wider support loops. Especially when considering how small this part is.

    The alternate option would be to increase the number of segments in the curve but that only makes sense if the model will be used for something other than baking. At which point, if shape accuracy is the primary concern, it may make more sense to look into alternate poly modeling or parametric modeling workflows. Which will be slightly easier to manage than traditional subdivision workflows.

    It really just depends on what the model will be used for.

  • Joao Sapiro
  • Kanni3d
    Offline / Send Message
    Kanni3d ngon master

    You've set the topology right here, it's just the trigger guard itself needs more base resolution to support that merge. Which is essentially only four polygons worth.


    The curved/bulge part (red) is pinching since it's not planar (blue). Add some loops/base resolution and you'll see an improvement.


  • naman
    Offline / Send Message
    naman node

    Thanks @FrankPolygon and @Kanni3d for your help. I am able to make it right now. 😄 Thanks for sharing the technique and methods.

  • Soldeus
    Offline / Send Message
    Soldeus polycounter lvl 10

    Hi! I've modeled this control panel piece that needed to have some specific button placement. I managed to achieve a working result without any noticeable shading errors at the distance/target we are aiming for, but, while it works, Im not happy at all with how the topology turned out and was wondering how to improve upon this! ( Problem was that this panel has a slight curve in 2 axis). Is this a case of just start with more topology so I could use the existing loops as supporting loops? Or is there any cleverer way to improve this model? Thanks in advance!!



  • Deqa
    Offline / Send Message
    Deqa polycounter lvl 8

    I'm stumped with this hood light and its Bevel transitions. Especially what appears to be a triangle termination around a corner. Am I approaching this incorrectly? I don't like how the bevel goes around the edge like this. I am certainly confused on how you could make certain edges of the hood light go from soft to sharp without intersecting bevels and producing pinched geometry, because some of the edges of the hood light are soft and then other parts are sharp.

    Am I wrong for thinking this could be done in one Bevel modifier? Even if I try to set up separate weights to create a soft edge in one area and a hard edge in another, it seems like an impossible task via weights or vertex groups. In addition, is the trim border a floater? My attempts at extruding that trim line and applying edge loops or bevels to it to get it to appear like a popped out trim have failed, needless to say. This seems like an easy task but I am stumped in more ways than one (mainly because I am staunchly thinking this could be done in a single Bevel modifier with Sharpen + Weights) and yet I end up with areas that need to be controlled more softly and some harder.


  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter

    @Deqa The edges on the model look a bit too sharp when compared to what's shown in the reference images. Though the fender is mostly flat surfaces, there are a number of angled transitions that are missing. Some of which are difficult to see. Gathering some higher quality references, with better lighting and more contrast, would make it easier to see exactly what's going on with the shapes.

    As far as the topology goes, though it possible to add all of the smaller scale support loops in a single operation, it will be much easier to do so once all of the larger shapes are fully blocked out. It's also completely acceptable to use support loops on just one side of a soft shape transition or to slide some of the double loops further apart to ease over the rounded edges. Just be sure to constrain any movement to the underlying surface. That way everything stays co-planar.

    Looking at some additional reference photos: the fenders and trim line do appear to be single piece molded fiberglass. Something that should probably be added to the block out. Especially before generating the smaller scale support loops. Getting all of the shapes into the block out and figuring out the basic loop flow will make it a lot easier to decide which edges need support loops on one side and which edges need support loops on both sides.

    Here's a quick example of what the rough block out could look like. Larger curves, like around the corners of the wheel well, can usually be added with single support loops. After all of the major shapes are present then the rest of the smaller support loops can be added with a bevel / chamfer operation.

    Though there's a number of different ways to setup variable width loop generation, sometimes it's just easier to apply the modifier and slide the support loops away from the edges that need to be softened. Really just depends on what makes the most sense for that specific part of the model.

    @Soldeus Are there any specific areas that seem like they could be improved or concerns about the results? A bit more context on the intended use case would also make it easier to give actionable feedback.

  • sacboi
    Offline / Send Message
    sacboi high dynamic range

    Unless you've an aversion to utilising destructive workflows I'd simply model that shape conventionally, by just using control/support loops whilst appropriately terminating poly strip transitions which really should be a straightforward process in of itself since those surfaces seem to appear planar or close too?!

    Edit:

    Haha...nerfed too the draw at a mere 60s 😀

    OP! Frank nailed it yet again so I'll advise following his example.

  • sacboi
    Offline / Send Message
    sacboi high dynamic range

    [double post deleted - still hung-over from NYE festivities 😎]

Sign In or Register to comment.