@macaron10 Looks like the basic topology flow should be workable. Running the support loops around the perimeter of the shapes will sharpen the outer profile but redirecting the loops across the flared cap will produce a softer, rounder profile at the end of the quillion. The blade and hilt sides of the cross guard aren't perfectly symmetrical in the reference image. So adjust the basic shape first, until everything lines up with the reference, then add the support loops with a single bevel operation.
@count23 If the edges aren't creased then it seems like the smoothing artifact is likely caused by the support loop that's directly behind the spherical cap. Uneven segment spacing near curved surfaces can cause unintended pinching artifacts when subdivision smoothing is applied. Moving that edge loop away from the shape transition should resolve the artifact.
Although it may not be necessary, it should be possible to simplify the mesh a bit further and that will preserve the segment spacing on the wall of the larger cylinder. Which will help reduce the possibility of smoothing artifacts appearing there. The example below shows what this could look like.
@kuronekoshiii The mesh in the original post looks like it should be fairly straight forward to remodel but if the mesh provided by the client is a lot more complex then it may make sense to take a slightly less conventional approach. If none of the proposed solutions will work within the project's constraints: another option would be to redirect the loop flow, as shown below, then manually move the highlighted edges in along the normals until the subdivided surface evens out.
There isn't a lot of geometry to work with. So, the results of manually compensating for the mismatched segments won't be perfect but that trade off might make sense if the only alternative is scrapping a complex mesh. If the customer expects artifact free subdivision and the supplied mesh needs to be re-worked then that's something they should probably be charged for.
Thanks for the reply. I think the bit that sticks me TBH is how to effectively cap the end of that cylinder while maintaining the proper profile of the cylinder itself. Is that all done by hand? Or did you blend a quad sphere with the end of a cylinder? the main reason i didnt try to cylindrically cap the end of my bay was because i couldn't effectively marry a quad sphere to the cylinder i was using. Part of the reason why my shape is so off is because i originally created the perpendicular circle to the cylinder, then quadded up the two loops and added some dividing edge loops.
@count23 When the number of segments in a cylinder is the limiting factor, one option for capping it with a quad sphere is to create a cube then use loop cut to match the number of segments and To Sphere to generate the shape. Scale the new quad sphere to fit then delete any sections that aren't needed or don't line up with the cylinder and use bridge edge loops to join the sphere to the end of the cylinder. The example below shows what this process could look like.
For this particular shape, blending a quad sphere to the open end of the cylinder is a viable option but it does tend to produce a bit more complexity than is necessary. An alternate approach is to use the cylinder's existing geometry to create a hemispherical quadrant.
This can be done by revolving the portion of the cylinder that matches the cut out then extruding and rotating the remaining edge section so it can be scaled up to match the missing curvature. The edge of the non-manifold area can be selected, extruded and rotated into position then connected to the rest of the shape using bridge edge loops.
Below is an example of what this process could look like. All of the geometry here is created and shaped using tools that produce consistent results when lofting or blending edge segments. The final mesh is a bit simpler than a standard quad sphere and most of the edges should already line up with the cut out in the references. Which makes it a lot easier to cut in additional details, without having to manually redirect the loop flow.
There's a number of different ways to approach these kinds of shapes but it's generally considered best practice to rely on tools and primitives to generate accurate geometry. So try to avoid manually extruding and moving individual edge segments to create complex, compound curves.
A lot of air-frame and hull forms are a series of lofted curves so there's always situations where raw primitives won't match. In these cases it can be helpful to keep the base mesh fairly simple and rely on subdivision to smooth out the shapes and add the geometry required to support the smaller details.
Thanks, that really helps clear things up, the approach you described is a lot easier than linking a quad sphere. I tried that after your first response and while it achieved the results, it was very limited (for instance, in the actua model i'm copying, since the cylinder is conical, that means the hemisphere needs to follow that angle to flatten out, a quad sphere didnt' acheive that).
Just a quick question on the gif you sent of your approach. So the Extrude/rotate and scale y+z you proposed went by a little fast. If i understand this correctly, what you're advising to do is extrude the edges you highlighted on an equivilent shape, then rotate down on the X axis using the original row as an origin to approximately halfway between the two edges i want to bridge, then use scaling on Y and Z to bring the extruded curve into the final right position?
Hmm interesting.. tried to remodel the original approach with the side view and.. doesn't get that bulgde... (ignore the none roundness and the little other artefact 😉)
@count23 Correct. Extruding and rotating the new edge into place ensures that the longitudinal edges remain parallel, until they are joined with the surrounding geometry. Constraining the scale operation by length and height ensures the width remains consistent and this helps prevent unintended surface deformation.
Without getting too far into the technical details, Catmull–Clark subdivision smooths by averaging the surface and this comes with some inherent accuracy limitations. Substantial shape and topology issues are usually easy to spot, when they produce obvious surface deformation, but abrupt changes to the surface shape, segment spacing and topology flow of the mesh can also produce subtle inconsistencies. Which affects the quality of the edge highlights and surface reflections.
These subtle types of surface artifacts can be difficult to spot and matte viewport materials tend to obscure both major smoothing and minor flow artifacts when modeling but the issues will show up when baking normals or using high gloss materials with sharp reflections. So, it's generally considered best practice to use a view-port material with bright highlight and wide roll-off. Materials with sharp highlights and surface reflections can also be used to evaluate surface quality. It's also considered best practice to do a detailed flow check of all areas that need high quality surfaces. Disruption or turbulence in the reflection map's lines generally coincides with surface quality issues.
Here's an approximation of what the original topology looks like when viewed with a variety of different view-port materials. The subtle shading artifact around the shape transition is still visible, it's just harder to see without rotating the highlight or compressing the levels in Photoshop. Smoothing and flow artifacts are lot easier to identify when using a more reflective material. Deflection and swirling in the reflection map shows where the changes in segment spacing generate turbulence in the surface flow.
In the original question, the second wire frame shows a slightly different topology layout that produces a stronger smoothing artifact and noticeable flow disruption. The increased flow deflection around the added support loops does suggest that moving the rear support loop away from the shape transition and routing the topology flow in a more consistent pattern should resolve the visible smoothing artifact.
Going back to the original example, moving the rear support loop away from the shape transition produces a more consistent segment spacing. Which reduces the turbulence in the reflection map and the visibility of the smoothing artifact. Pushing the support loop closer to the shape transition has the opposite effect. While the result isn't perfect, moving the rear support loop should resolve the smoothing artifact near the shape transition.
Maintaining a relatively consistent segment spacing and simplifying the topology routing produces a much smoother surface flow and shape transition.
Below is a comparison of the different segment spacing and topology layouts. Using more geometry can make it easier to add details to a contiguous mesh but there's also diminishing returns and after a certain point it doesn't make sense to keep re-working a mesh that's good enough.
The hull in the reference images looks like it has a matte finish and minimal surface details. If the player or viewer isn't close enough to see any minor artifacts then there's usually going to be minimal benefit to improving the results beyond what's required for the intended use case for the asset.
Another thing to consider is that the hull form is a series of complex, compound curves with a lot of intersecting shapes. So, it probably makes sense to block out all of the larger forms first then work through another block out pass and add all of the smaller details like the landing bay doors. If most of the surface details like panel lines, greebles, windows, etc. will be added with a texture then the base mesh can be kept fairly simple.
Avoiding unnecessary complexity will make the modeling process a lot easier and can help prevent surface flow issues and visible smoothing artifacts. Below is an example of how consistent segment spacing and simplified topology can be used to create the basic shapes while also minimizing flow disruptions.
Recap:
Use a viewport material with a visible highlight or reflection to evaluate surface quality while modeling.
Flow check parts that require high quality surface reflections.
I was refering to the OP's question about: Am I on the right track and just need to do a few tweaks? And was also somekind wondering why my simplest re-modelling (the same topology) doesn't have such a clearly visible buldge.. so it could be enhanced in a not too cumbersome manner.. The model also doesn't have such a shiny material.. the reference even has some grungy textures...
Of course the master of it all (bowing head to @FrankPolygon .. i'm unworthy) can easily show different approaches which end up in much beter results.
"Avoiding unnecessary complexity will make the modeling process a lot easier and can help prevent surface flow issues and visible smoothing artifacts."
Indeed!
For me, a very hard lesson too learn among others.
Thanks for the help, this ended up beving my "current" final shape for the area in the end
Seems to have turned out pretty good. I am using shrinkwrap to confirm any deltas in the area onto a version of the mesh without the cutout. Definitely feels easier to maintain now compared to my original approach and gave a result with no "dimples" in the shading. I'm still a little fuzzy on the appropriate constraining when extruding the loop to meet the revolved section, but that'll just be a bit of practice i guess.
EDIT: looking at the screenshot now nad the subd cage, I may have misinterpreted the constrain axis and should have scaled X closer to 0 to keep the longitudinal rows straight, if i understand frank's explanation earlier correctly.
Hi Guys .. 1st post here ... I hope I'm using the thread correctly. I'm pretty new to SubD hard surface modeling I need help with the corners of a sci-fi panel cutout. Any help would be great .. I'm getting brain meltdown deleting and redrawing edges. Anyway thanks for the noob help.
@dunnie How does it appear when you sub-divide? Where are the problematic areas? Topology seemingly looks fine, but once it artifacts/pinches, it'll be easier to determine what topology will work.
Guys - First of all Thanks SO much for the help ... the issue is that it is on a slightly curved piece of geo ..I am going to try wirrexx's solution. Trying to keep this light because its part of a larger spaceship,
Attatched is an OBJ of the barrel panel I am working on. Its curved so any thoughts will help.
This piece makes part of the Corinthian roman column I am making. I know this should have been an easy problem to solve, but I am too dumb to figure this out.
I want to see if it is possible to fix this instead of re-doing it.
As it's curved, you then need to have evenly spaced geometry, and plenty of it (the more you have, the more natural support loops they provide). Avoid having poles on corners, as that'd create a pinch.
Hello, I discovered this thread after tons of time spending on searching ways for my... "complex problem", just as I thought. So, one what I know, is just that I want to connect these two objects and filet their edges between their sides. I want reach exactly what @ZacD showed here in 89 page, but I don't know how do even start (how to "extract"...), which modifiers use and so on. It seems that this thread is my last hope.
Hello, I discovered this thread after tons of time spending on searching ways for my... "complex problem", just as I thought. So, one what I know, is just that I want to connect these two objects and filet their edges between their sides. I want reach exactly what @ZacD showed here, but I don't know how do even start (how to "extract"...), which modifiers use and so on. It seems that this topic is my last hope.
@Varravik it's worth keeping in mind ZacD had shared his solution to enable 'joining' two separate objects modeled with varying number of edge segments, basically faking a smooth transition using 'floaters' (floating geometry) in order to bake detail without generating errors from a high poly mesh too low poly proxy, then to be eventually rendered out in a realtime engine.
So I'm just wondering if your goal is to create a game asset or high resolution 3D model plus whether also relates to converting CAD/Nurbs data too polygonal?
"I just extract a loop of polygons that go around the cylinder, then I push one of the edges out and flatten the shape so it looks like a flat doughnut with the inner loop matching the original mesh it was extracted from, then I use a shrink wrap deformer and use it on the gun barrel. I then extrude the inner loop up, push the shape a bit so it's not intersecting, and your done. You can also model it in a way where the floating loop is connected to one of the meshes, or you can shrink wrap to both surfaces if you need to."
The ridges disrupt the spacing of curve's segments and this causes smoothing artifacts when subdivided. There's a few different ways to resolve these types of artifacts. Deciding which approach to use really depends on what the model will be used for and how accurate the surface needs to be.
Compensating for unintended deformation by moving the geometry back into alignment with the curve's arc can help reduce the visibility of some smoothing artifacts but might not be able to completely resolve all of them. Using scale or shrink wrap on the highlighted edges would be a quick and easy way to achieve usable, though slightly imperfect, results with the existing mesh.
Equalizing the segment spacing and rotating the mesh, so the ridges fall be the existing geometry in the curve, makes it easier to simplify the mesh without producing obvious smoothing artifacts. In some cases it may be necessary to adjust the position of the highlighted edges but most of the time the mesh will subdivide cleanly if the edge segments of the curve are relatively consistent.
Increasing the number of segments in the curve will generally increase the accuracy of the subdivided surface and can help reduce smoothing artifacts but does tend to reduce the overall editability of the mesh. Deciding to place the ridges between or on top of the vertical edge segments really just depends on whether or not the added complexity helps control the shapes.
Arbitrarily increasing the geometry density, beyond what's required to achieve clean results at the intended view distance, can be counter productive because it tends to introduce unnecessary complexity that makes it difficult to edit the mesh. It can also constrict the maximum width of edge highlights, when using the existing geometry of a curved surface as support loops for attached shapes.
Floating geometry can also be used to add surface details without increasing mesh complexity but its use tends to be limited to high poly meshes for baking. It can also be difficult to avoid visible seams around complex shape intersections when using lots of floating elements.
Recap: Uneven segment spacing can cause smoothing artifacts on curved surfaces. Adjusting the geometry to match the curvature when subdivided can help reduce smoothing artifacts but may not fully resolve them. Aligning the segments of curves with surface details allows the existing geometry to be used as support and can help maintain consistent segment spacing. Increasing the amount of geometry in a curved surface can increase the accuracy when subdivided but also tends to reduce the overall editability of the mesh. Evaluate how the mesh will be used and how closely it will be viewed and use an appropriate amount of geometry that subdivides cleanly and remains easy to edit.
@sacboi thank you for respond. However, I guess that this way with cylinder is not enough for my complex case. It is very parametrical surface, and probably badly maked, just look.
Trying to 3D model a skillet, but there's this ugly shading issue I can't seem to solve when connecting the handle to the rest of the skillet. Below I also attached the topology.
@Varravik can you please share a screenshot of the object you're modeling whether background image - reference...etc or even a brief description because it's really for me at least, unclear what you're trying to do?
Aside from your support loops being far too tight - you are clustering a lot of topology in such a small area, with sparse surrounding geo, giving you a pinch. Basically detailling too soon.
Try having much more geometry/curvature established on the grip of the gun before trying to connect the trigger guard.
@SkinnyM Have to agree with Eric's recommendation to block out the major shapes before merging them and adding support loops.
To add to what's already been said: blending complex shape intersections is all about creating accurate shapes that have relatively consistent segment spacing. Using more geometry can help with certain types of smoothing artifacts but also makes it difficult to adjust the shapes. This is why it's a lot easier to resolve major topology flow issues during the block out, without all the added complexity of having to manage blending the support loops into the shapes.
Focus on creating accurate shapes first then solve the topology flow issues around the shape intersections by matching the segments of adjacent shapes and using the existing geometry as support. After the basic topology routing is solved the secondary details can be added and sharped with support loops.
Below is an example of what this type of block out process could look like. Working through the shape accuracy and topology flow issues first allows the the shapes to define the support loops, rather than the support loops limiting the accuracy of the shapes.
It's also important to minimize the amount of surface deformation generated by merging or reducing unneeded geometry. Try to avoid making extreme changes in the underlying curvature of the intersecting shapes. Significant differences between the intersecting geometry can usually be constrained to the area between the support loops or averaged out over a larger surface.
Recap: Us the block out process to generate accurate shapes and solve topology flow issues. Use the minimum amount of geometry required to accurately represent the shapes and try to maintain relatively consistent segment spacing along curves. Match the number of segments in adjacent features and use the existing geometry as support for shape intersections. Let the major forms guide the support loop flow.
I started to follow this thread a few years ago, but I decided to register here because I have just run into a problem that I can't solve alone. I learn a lot from this thread about hard surface modelling and I don't have problem with mechanical objects but I challenged myself with an TetraPak milk carton box and it just don't want to look like as I want it to be :( . I can feel approximately where the problem is but I don't know how to approach it. My problem I think is the varying bevel along the edges and parts were sharp folding occur. Maybe I could model the folding separately? Or should I create the base box first and extrude ? Or use manual topology ? Here is one of my attempts and a reference image.
This is the complicated part of my model that I am facing a lot of issue with. I tried several different methods to fix this but none of them are working out correctly for me. The peculiar thing about this shape is that while it looks like a sphere, the edges are going inside like a C shape instead of being round. Also note that the size of the holes on the left and right is different from the holes in which the dials are placed.
This is one attempt in which I tried to boolean a sphere with a cube and then cut the holes out of it but this is giving artifacts because of the the holes being cut at random places (I have tried fixing the topology but that doesn't really solve it entirely) and also, it is round instead of the pinching shape the reference has (see what I mean in the next images).
I want the flow to match the red curve but most solutions that I can think of are giving me a flow similar to the yellow curve.
@parikshittiwari Regarding the flow of the red curve, I found some real gun scope references on Amazon and the shape seems similar to what you've created using the Sphere and Boolean workflow. Although, if you look at the middle image it seems to give the illusion of the red curve flow but it could've been exaggerated a little bit in the concept. It should be fine as long as you follow a real life reference in this case.
Thanks a lot for searching for the references and I really didn't think it could just be an illusion. The shape itself was very tricky precisely because I was being meticulous over the flow but you solved that for me.
The YouTube video also teaches an extremely useful trick so thanks for that too, much appreciated.
If you look at the area marked in red, it basically has a cylindrical shape with one cylinder(green light one) inside another cylinder(the silver rail containing the lights) and then another bigger cylindrical shape at the bottom holding it together with the top square part. I can't see to get the scale right for this entire part here.
If I try to follow only the silver cylinder part and adjust scale then I can't seem to find a way to add the black cylindrical plating at the bottom of it since it feels too small.
If I make something really big to fill up the entire space then the edges of the entire middle body crosses this section of the body which doesn't seem to be matching the reference so I am a little confused as to what I should be doing here.
Soft hard surface objects can be tricky to model but taking apart one of these packages or gathering references of them disassembled can provide some insight into how the paper is cut and folded to create the carton. Analyzing how these pieces fit together will make it easier to come up with a plan for modeling the shapes.
While there's a few different ways to approach this, depending on the level of accuracy required, it's generally a good idea to start by blocking out the shapes. Keeping things fairly simple during the block out makes it easier to adjust the larger forms and change the topology flow. Try to resolve any major shape or topology flow issues before adding the support loops.
If simplified fold details are acceptable then one of the easier modeling strategies would be to cut in the basic shape of the flaps and sealed seam then inset or extrude the outline to create some depth. While this approach is fairly quick and detailed enough for most background props, it won't be completely accurate. Which could be an issue if the model needs to be viewed up close.
If more complex fold details are required then it probably makes sense to model the entire seam standing up then fold it flat against the carton. This approach is a bit more involved but produces more accurate fold details. Below is an example of what this modeling process could look like.
Start by establishing the overall form then apply a mirror modifier and add a central loop cut. Select the outside edges and run a chamfer operation to create the fold line around the corners of the packing. Cut in an edge loop to define the basic shape of the flaps. Extrude this section and merge the geometry into a point. Use a bevel operation to create a square edge, that's the width of the sealed seam, along the central edge loop. Extrude the new loops upwards to create the standing seam. Add the support loops for the outer perimeter of the carton and additional loops for deforming the sides. Select the standing seam and fold it over then fold in the corners.
Adjust the position of the loops around the top and bottom of the carton. Looking at the references: the top tends to bow inwards and the bottom tends to bow outwards slightly. Loop select the edges that define the perimeter of the folded seam and corners of the box and add support loops with a bevel / chamfer operation. The folded seam on the bottom of the carton is the same as the top, only the flaps are folded inwards instead of outwards.
Though this mesh is water tight, the area under the folded seams will have some potential for overlap and could be further simplified by merging it into the underlying surface. In most cases this shouldn't be required but it is an option for any sort of edge case where the overlap causes issues.
Recap: Gather additional reference information about the folded parts of the object and use that to develop a plan for modeling the shapes. Block out the major forms and resolve the topology flow issues before adding support loops. Try to keep the basic geometry fairly simple, with relatively consistent segment spacing, then let the subdivision do most of the smoothing work around the corners of the folded flaps.
Lots of good advice on how to approach modeling the truncated sphere but when shape accuracy is important it's often beneficial to rely on mathematically consistent primitive geometry as a starting point. There's also nothing inherently wrong with using a sphere that has a central pole. It just requires a bit of adjustment to find the right number of rings and segments to get the geometry to line up with the section planes. Below is an example of what this process and resulting topology could look like.
There's also some situations where it makes more sense to use a regular sphere instead of a quad sphere or icosphere, since it's a lot easier to adjust the number of segments to line up with intersecting geometry. The remainder of partial rings can also be used as support loops. Without having to add additional geometry that would otherwise disturb the segment spacing and quality of the surface.
Quad sphere geometry tends to work well with symmetrical section planes but can run into issues when blending asymmetrical shape intersections. Both topology layouts are viable and deciding which to use really depends on what the adjacent geometry looks like.
For your latest shape question: the silhouettes on the reference sheet and the detailed side views seem to show a cut out between the two cylinders. Try using the two, parallel cylinders ahead of that area as a proportional reference then block out the larger cylinder towards the back and see how it looks with the gap that appears between the top and bottom.
I think the answer (as always) is in the realm of "moar geo". If you are modelling an (extreme) closeup of any object with any kind of detailed appearance (especially with sharp, uneven changes, like these sharp folds) you need enough geo to support the attempted level of detail. If you are doing this for like a high quality extreme close up for a commercial, the model you show is very low poly as well as uneven. If it's a game object that will only be seen from a distance, it's fine or a bit too detailed (it could literally be 6 polygons and a texture/normal map). If you want to catch the kind of details shown in the photograph, you need pretty extreme high amounts of geo to hold those subtle deformations. Probably a combination of modelling the basic folds and sculpting the details would be ideal in that case.
Replies
But if the payment is low, remodeling is not a good idea 😥
But if the payment is low, remodeling is not a good idea 😥
@macaron10 Looks like the basic topology flow should be workable. Running the support loops around the perimeter of the shapes will sharpen the outer profile but redirecting the loops across the flared cap will produce a softer, rounder profile at the end of the quillion. The blade and hilt sides of the cross guard aren't perfectly symmetrical in the reference image. So adjust the basic shape first, until everything lines up with the reference, then add the support loops with a single bevel operation.
@count23 If the edges aren't creased then it seems like the smoothing artifact is likely caused by the support loop that's directly behind the spherical cap. Uneven segment spacing near curved surfaces can cause unintended pinching artifacts when subdivision smoothing is applied. Moving that edge loop away from the shape transition should resolve the artifact.
Although it may not be necessary, it should be possible to simplify the mesh a bit further and that will preserve the segment spacing on the wall of the larger cylinder. Which will help reduce the possibility of smoothing artifacts appearing there. The example below shows what this could look like.
@kuronekoshiii The mesh in the original post looks like it should be fairly straight forward to remodel but if the mesh provided by the client is a lot more complex then it may make sense to take a slightly less conventional approach. If none of the proposed solutions will work within the project's constraints: another option would be to redirect the loop flow, as shown below, then manually move the highlighted edges in along the normals until the subdivided surface evens out.
There isn't a lot of geometry to work with. So, the results of manually compensating for the mismatched segments won't be perfect but that trade off might make sense if the only alternative is scrapping a complex mesh. If the customer expects artifact free subdivision and the supplied mesh needs to be re-worked then that's something they should probably be charged for.
Thanks for the reply. I think the bit that sticks me TBH is how to effectively cap the end of that cylinder while maintaining the proper profile of the cylinder itself. Is that all done by hand? Or did you blend a quad sphere with the end of a cylinder? the main reason i didnt try to cylindrically cap the end of my bay was because i couldn't effectively marry a quad sphere to the cylinder i was using. Part of the reason why my shape is so off is because i originally created the perpendicular circle to the cylinder, then quadded up the two loops and added some dividing edge loops.
@count23 When the number of segments in a cylinder is the limiting factor, one option for capping it with a quad sphere is to create a cube then use loop cut to match the number of segments and To Sphere to generate the shape. Scale the new quad sphere to fit then delete any sections that aren't needed or don't line up with the cylinder and use bridge edge loops to join the sphere to the end of the cylinder. The example below shows what this process could look like.
For this particular shape, blending a quad sphere to the open end of the cylinder is a viable option but it does tend to produce a bit more complexity than is necessary. An alternate approach is to use the cylinder's existing geometry to create a hemispherical quadrant.
This can be done by revolving the portion of the cylinder that matches the cut out then extruding and rotating the remaining edge section so it can be scaled up to match the missing curvature. The edge of the non-manifold area can be selected, extruded and rotated into position then connected to the rest of the shape using bridge edge loops.
Below is an example of what this process could look like. All of the geometry here is created and shaped using tools that produce consistent results when lofting or blending edge segments. The final mesh is a bit simpler than a standard quad sphere and most of the edges should already line up with the cut out in the references. Which makes it a lot easier to cut in additional details, without having to manually redirect the loop flow.
There's a number of different ways to approach these kinds of shapes but it's generally considered best practice to rely on tools and primitives to generate accurate geometry. So try to avoid manually extruding and moving individual edge segments to create complex, compound curves.
A lot of air-frame and hull forms are a series of lofted curves so there's always situations where raw primitives won't match. In these cases it can be helpful to keep the base mesh fairly simple and rely on subdivision to smooth out the shapes and add the geometry required to support the smaller details.
Thanks, that really helps clear things up, the approach you described is a lot easier than linking a quad sphere. I tried that after your first response and while it achieved the results, it was very limited (for instance, in the actua model i'm copying, since the cylinder is conical, that means the hemisphere needs to follow that angle to flatten out, a quad sphere didnt' acheive that).
Just a quick question on the gif you sent of your approach. So the Extrude/rotate and scale y+z you proposed went by a little fast. If i understand this correctly, what you're advising to do is extrude the edges you highlighted on an equivilent shape, then rotate down on the X axis using the original row as an origin to approximately halfway between the two edges i want to bridge, then use scaling on Y and Z to bring the extruded curve into the final right position?
Hmm interesting.. tried to remodel the original approach with the side view and.. doesn't get that bulgde... (ignore the none roundness and the little other artefact 😉)
now put on a shiney material and see what happens
@count23 Correct. Extruding and rotating the new edge into place ensures that the longitudinal edges remain parallel, until they are joined with the surrounding geometry. Constraining the scale operation by length and height ensures the width remains consistent and this helps prevent unintended surface deformation.
Without getting too far into the technical details, Catmull–Clark subdivision smooths by averaging the surface and this comes with some inherent accuracy limitations. Substantial shape and topology issues are usually easy to spot, when they produce obvious surface deformation, but abrupt changes to the surface shape, segment spacing and topology flow of the mesh can also produce subtle inconsistencies. Which affects the quality of the edge highlights and surface reflections.
These subtle types of surface artifacts can be difficult to spot and matte viewport materials tend to obscure both major smoothing and minor flow artifacts when modeling but the issues will show up when baking normals or using high gloss materials with sharp reflections. So, it's generally considered best practice to use a view-port material with bright highlight and wide roll-off. Materials with sharp highlights and surface reflections can also be used to evaluate surface quality. It's also considered best practice to do a detailed flow check of all areas that need high quality surfaces. Disruption or turbulence in the reflection map's lines generally coincides with surface quality issues.
Here's an approximation of what the original topology looks like when viewed with a variety of different view-port materials. The subtle shading artifact around the shape transition is still visible, it's just harder to see without rotating the highlight or compressing the levels in Photoshop. Smoothing and flow artifacts are lot easier to identify when using a more reflective material. Deflection and swirling in the reflection map shows where the changes in segment spacing generate turbulence in the surface flow.
In the original question, the second wire frame shows a slightly different topology layout that produces a stronger smoothing artifact and noticeable flow disruption. The increased flow deflection around the added support loops does suggest that moving the rear support loop away from the shape transition and routing the topology flow in a more consistent pattern should resolve the visible smoothing artifact.
Going back to the original example, moving the rear support loop away from the shape transition produces a more consistent segment spacing. Which reduces the turbulence in the reflection map and the visibility of the smoothing artifact. Pushing the support loop closer to the shape transition has the opposite effect. While the result isn't perfect, moving the rear support loop should resolve the smoothing artifact near the shape transition.
Maintaining a relatively consistent segment spacing and simplifying the topology routing produces a much smoother surface flow and shape transition.
Below is a comparison of the different segment spacing and topology layouts. Using more geometry can make it easier to add details to a contiguous mesh but there's also diminishing returns and after a certain point it doesn't make sense to keep re-working a mesh that's good enough.
The hull in the reference images looks like it has a matte finish and minimal surface details. If the player or viewer isn't close enough to see any minor artifacts then there's usually going to be minimal benefit to improving the results beyond what's required for the intended use case for the asset.
Another thing to consider is that the hull form is a series of complex, compound curves with a lot of intersecting shapes. So, it probably makes sense to block out all of the larger forms first then work through another block out pass and add all of the smaller details like the landing bay doors. If most of the surface details like panel lines, greebles, windows, etc. will be added with a texture then the base mesh can be kept fairly simple.
Avoiding unnecessary complexity will make the modeling process a lot easier and can help prevent surface flow issues and visible smoothing artifacts. Below is an example of how consistent segment spacing and simplified topology can be used to create the basic shapes while also minimizing flow disruptions.
Recap:
Use a viewport material with a visible highlight or reflection to evaluate surface quality while modeling.
Flow check parts that require high quality surface reflections.
Block out the larger forms then add the details.
I was refering to the OP's question about: Am I on the right track and just need to do a few tweaks? And was also somekind wondering why my simplest re-modelling (the same topology) doesn't have such a clearly visible buldge.. so it could be enhanced in a not too cumbersome manner.. The model also doesn't have such a shiny material.. the reference even has some grungy textures...
Of course the master of it all (bowing head to @FrankPolygon .. i'm unworthy) can easily show different approaches which end up in much beter results.
"Avoiding unnecessary complexity will make the modeling process a lot easier and can help prevent surface flow issues and visible smoothing artifacts."
Indeed!
For me, a very hard lesson too learn among others.
Thanks for the help, this ended up beving my "current" final shape for the area in the end
Seems to have turned out pretty good. I am using shrinkwrap to confirm any deltas in the area onto a version of the mesh without the cutout. Definitely feels easier to maintain now compared to my original approach and gave a result with no "dimples" in the shading. I'm still a little fuzzy on the appropriate constraining when extruding the loop to meet the revolved section, but that'll just be a bit of practice i guess.
EDIT: looking at the screenshot now nad the subd cage, I may have misinterpreted the constrain axis and should have scaled X closer to 0 to keep the longitudinal rows straight, if i understand frank's explanation earlier correctly.
thanks
Hi Guys .. 1st post here ... I hope I'm using the thread correctly. I'm pretty new to SubD hard surface modeling I need help with the corners of a sci-fi panel cutout. Any help would be great .. I'm getting brain meltdown deleting and redrawing edges. Anyway thanks for the noob help.
@dunnie How does it appear when you sub-divide? Where are the problematic areas? Topology seemingly looks fine, but once it artifacts/pinches, it'll be easier to determine what topology will work.
I feel as if I have seen this post before... And Also reacted to it. Because this one Is a Deja Vu for me.
If it's a flat surface, you can end on ngons and do whatever you want. it will not be visible.
Guys - First of all Thanks SO much for the help ... the issue is that it is on a slightly curved piece of geo ..I am going to try wirrexx's solution. Trying to keep this light because its part of a larger spaceship,
Attatched is an OBJ of the barrel panel I am working on. Its curved so any thoughts will help.
Hello everyone!
Can someone help me with this object?
This piece makes part of the Corinthian roman column I am making. I know this should have been an easy problem to solve, but I am too dumb to figure this out.
I want to see if it is possible to fix this instead of re-doing it.
Thank you in advance!
And sorry for bothering you. 😔
As it's curved, you then need to have evenly spaced geometry, and plenty of it (the more you have, the more natural support loops they provide). Avoid having poles on corners, as that'd create a pinch.
Hello, I discovered this thread after tons of time spending on searching ways for my... "complex problem", just as I thought. So, one what I know, is just that I want to connect these two objects and filet their edges between their sides. I want reach exactly what @ZacD showed here in 89 page, but I don't know how do even start (how to "extract"...), which modifiers use and so on. It seems that this thread is my last hope.
Hello, I discovered this thread after tons of time spending on searching ways for my... "complex problem", just as I thought. So, one what I know, is just that I want to connect these two objects and filet their edges between their sides. I want reach exactly what @ZacD showed here, but I don't know how do even start (how to "extract"...), which modifiers use and so on. It seems that this topic is my last hope.
@Varravik it's worth keeping in mind ZacD had shared his solution to enable 'joining' two separate objects modeled with varying number of edge segments, basically faking a smooth transition using 'floaters' (floating geometry) in order to bake detail without generating errors from a high poly mesh too low poly proxy, then to be eventually rendered out in a realtime engine.
So I'm just wondering if your goal is to create a game asset or high resolution 3D model plus whether also relates to converting CAD/Nurbs data too polygonal?
___________________________________________________________________________________________________________________________________________
@ZacD said
"I prefer to model the "weld" as a separate floating piece of geometry."
https://polycount.com/discussion/comment/2204079/#Comment_2204079
@ZacD said
"I just extract a loop of polygons that go around the cylinder, then I push one of the edges out and flatten the shape so it looks like a flat doughnut with the inner loop matching the original mesh it was extracted from, then I use a shrink wrap deformer and use it on the gun barrel. I then extrude the inner loop up, push the shape a bit so it's not intersecting, and your done. You can also model it in a way where the floating loop is connected to one of the meshes, or you can shrink wrap to both surfaces if you need to."
https://polycount.com/discussion/comment/2204378/#Comment_2204378
__________________________________________________________________________________________________________________________________________
And by way of explaining terms annotated to describe each step of his tutorial i.e.
Well, I kinda fixed the topology by deleting and re-doing the circular part.
Unfortunately, is very uneven, as shown in the 1st image! I will keep trying fixing it ;/
How do I manage the mesh to get those ridges on the bucket with no pinching
"How The F*#% Do I Model This? - Reply for help with specific shapes - (Post attempt before asking)"
@Lemenus firstly, let us have a peek at what you've done so far?
@anishraaj Welcome to Polycount. Consider checking out the forum information and introduction thread.
The ridges disrupt the spacing of curve's segments and this causes smoothing artifacts when subdivided. There's a few different ways to resolve these types of artifacts. Deciding which approach to use really depends on what the model will be used for and how accurate the surface needs to be.
Compensating for unintended deformation by moving the geometry back into alignment with the curve's arc can help reduce the visibility of some smoothing artifacts but might not be able to completely resolve all of them. Using scale or shrink wrap on the highlighted edges would be a quick and easy way to achieve usable, though slightly imperfect, results with the existing mesh.
Equalizing the segment spacing and rotating the mesh, so the ridges fall be the existing geometry in the curve, makes it easier to simplify the mesh without producing obvious smoothing artifacts. In some cases it may be necessary to adjust the position of the highlighted edges but most of the time the mesh will subdivide cleanly if the edge segments of the curve are relatively consistent.
Increasing the number of segments in the curve will generally increase the accuracy of the subdivided surface and can help reduce smoothing artifacts but does tend to reduce the overall editability of the mesh. Deciding to place the ridges between or on top of the vertical edge segments really just depends on whether or not the added complexity helps control the shapes.
Arbitrarily increasing the geometry density, beyond what's required to achieve clean results at the intended view distance, can be counter productive because it tends to introduce unnecessary complexity that makes it difficult to edit the mesh. It can also constrict the maximum width of edge highlights, when using the existing geometry of a curved surface as support loops for attached shapes.
Floating geometry can also be used to add surface details without increasing mesh complexity but its use tends to be limited to high poly meshes for baking. It can also be difficult to avoid visible seams around complex shape intersections when using lots of floating elements.
Recap: Uneven segment spacing can cause smoothing artifacts on curved surfaces. Adjusting the geometry to match the curvature when subdivided can help reduce smoothing artifacts but may not fully resolve them. Aligning the segments of curves with surface details allows the existing geometry to be used as support and can help maintain consistent segment spacing. Increasing the amount of geometry in a curved surface can increase the accuracy when subdivided but also tends to reduce the overall editability of the mesh. Evaluate how the mesh will be used and how closely it will be viewed and use an appropriate amount of geometry that subdivides cleanly and remains easy to edit.
go Go go go Frankkky you are awesome! Love that you included Floaters!!
I tried to model this part without using boolean，but the result is not good and the process was difficult
Is there a better way to reach it? Or just use Boolean?
Forgive my bad topology, I'm newbie and a little eager to get things done😰
how does it smooth?
@sacboi thank you for respond. However, I guess that this way with cylinder is not enough for my complex case. It is very parametrical surface, and probably badly maked, just look.
Trying to 3D model a skillet, but there's this ugly shading issue I can't seem to solve when connecting the handle to the rest of the skillet. Below I also attached the topology.
use subdivisons, your shading can only do so much with so little vertices especially on curved surfaces/round objects.
@Varravik can you please share a screenshot of the object you're modeling whether background image - reference...etc or even a brief description because it's really for me at least, unclear what you're trying to do?
Thanks!
Hey everyone!
I'm trying to model this using subdiv.
And I'm stuck at this place:
This is the mesh:
And subdivided:
I can't fugire out how to get rid of these pinches: 😤
Any ideas how to solve it? 🙄
@SkinnyM
Aside from your support loops being far too tight - you are clustering a lot of topology in such a small area, with sparse surrounding geo, giving you a pinch. Basically detailling too soon.
Try having much more geometry/curvature established on the grip of the gun before trying to connect the trigger guard.
"you are clustering a lot of topology in such a small area" - that's because of support loops going from other places.
Ok, I'll try to finish the grip then attach the trigger guard. Thanks.
@SkinnyM Have to agree with Eric's recommendation to block out the major shapes before merging them and adding support loops.
To add to what's already been said: blending complex shape intersections is all about creating accurate shapes that have relatively consistent segment spacing. Using more geometry can help with certain types of smoothing artifacts but also makes it difficult to adjust the shapes. This is why it's a lot easier to resolve major topology flow issues during the block out, without all the added complexity of having to manage blending the support loops into the shapes.
Focus on creating accurate shapes first then solve the topology flow issues around the shape intersections by matching the segments of adjacent shapes and using the existing geometry as support. After the basic topology routing is solved the secondary details can be added and sharped with support loops.
Below is an example of what this type of block out process could look like. Working through the shape accuracy and topology flow issues first allows the the shapes to define the support loops, rather than the support loops limiting the accuracy of the shapes.
It's also important to minimize the amount of surface deformation generated by merging or reducing unneeded geometry. Try to avoid making extreme changes in the underlying curvature of the intersecting shapes. Significant differences between the intersecting geometry can usually be constrained to the area between the support loops or averaged out over a larger surface.
Recap: Us the block out process to generate accurate shapes and solve topology flow issues. Use the minimum amount of geometry required to accurately represent the shapes and try to maintain relatively consistent segment spacing along curves. Match the number of segments in adjacent features and use the existing geometry as support for shape intersections. Let the major forms guide the support loop flow.
Well, that should help a lot. Many thanks for the detailed answer!
just wanted to say: thank you @FrankPolygon its great that you put so much thought and work int your posts explaining this.
Hello guys,
I started to follow this thread a few years ago, but I decided to register here because I have just run into a problem that I can't solve alone. I learn a lot from this thread about hard surface modelling and I don't have problem with mechanical objects but I challenged myself with an TetraPak milk carton box and it just don't want to look like as I want it to be :( . I can feel approximately where the problem is but I don't know how to approach it. My problem I think is the varying bevel along the edges and parts were sharp folding occur. Maybe I could model the folding separately? Or should I create the base box first and extrude ? Or use manual topology ? Here is one of my attempts and a reference image.
My 20th attempt:
Some reference like picture:
Thanks!
laeion
This is the complicated part of my model that I am facing a lot of issue with. I tried several different methods to fix this but none of them are working out correctly for me. The peculiar thing about this shape is that while it looks like a sphere, the edges are going inside like a C shape instead of being round. Also note that the size of the holes on the left and right is different from the holes in which the dials are placed.
This is one attempt in which I tried to boolean a sphere with a cube and then cut the holes out of it but this is giving artifacts because of the the holes being cut at random places (I have tried fixing the topology but that doesn't really solve it entirely) and also, it is round instead of the pinching shape the reference has (see what I mean in the next images).
I want the flow to match the red curve but most solutions that I can think of are giving me a flow similar to the yellow curve.
@parikshittiwari, I created this quickly, I hope it can give an idea how to approach that shape:
@parikshittiwari Regarding the flow of the red curve, I found some real gun scope references on Amazon and the shape seems similar to what you've created using the Sphere and Boolean workflow. Although, if you look at the middle image it seems to give the illusion of the red curve flow but it could've been exaggerated a little bit in the concept. It should be fine as long as you follow a real life reference in this case.
Now, instead of the Boolean you can also try this https://www.youtube.com/watch?v=4ioxnMq-9io&t=4s
Hope this helps!
Thats a very good way of doing it. Thanks for the advice, really helps to understand the unique approach here.
Thanks a lot for searching for the references and I really didn't think it could just be an illusion. The shape itself was very tricky precisely because I was being meticulous over the flow but you solved that for me.
The YouTube video also teaches an extremely useful trick so thanks for that too, much appreciated.
Another question for my model here.
If you look at the area marked in red, it basically has a cylindrical shape with one cylinder(green light one) inside another cylinder(the silver rail containing the lights) and then another bigger cylindrical shape at the bottom holding it together with the top square part. I can't see to get the scale right for this entire part here.
If I try to follow only the silver cylinder part and adjust scale then I can't seem to find a way to add the black cylindrical plating at the bottom of it since it feels too small.
If I make something really big to fill up the entire space then the edges of the entire middle body crosses this section of the body which doesn't seem to be matching the reference so I am a little confused as to what I should be doing here.
@Neox Really appreciate the kudos.
@laeion Welcome to Polycount. Consider checking out the forum information and introduction thread.
Soft hard surface objects can be tricky to model but taking apart one of these packages or gathering references of them disassembled can provide some insight into how the paper is cut and folded to create the carton. Analyzing how these pieces fit together will make it easier to come up with a plan for modeling the shapes.
While there's a few different ways to approach this, depending on the level of accuracy required, it's generally a good idea to start by blocking out the shapes. Keeping things fairly simple during the block out makes it easier to adjust the larger forms and change the topology flow. Try to resolve any major shape or topology flow issues before adding the support loops.
If simplified fold details are acceptable then one of the easier modeling strategies would be to cut in the basic shape of the flaps and sealed seam then inset or extrude the outline to create some depth. While this approach is fairly quick and detailed enough for most background props, it won't be completely accurate. Which could be an issue if the model needs to be viewed up close.
If more complex fold details are required then it probably makes sense to model the entire seam standing up then fold it flat against the carton. This approach is a bit more involved but produces more accurate fold details. Below is an example of what this modeling process could look like.
Start by establishing the overall form then apply a mirror modifier and add a central loop cut. Select the outside edges and run a chamfer operation to create the fold line around the corners of the packing. Cut in an edge loop to define the basic shape of the flaps. Extrude this section and merge the geometry into a point. Use a bevel operation to create a square edge, that's the width of the sealed seam, along the central edge loop. Extrude the new loops upwards to create the standing seam. Add the support loops for the outer perimeter of the carton and additional loops for deforming the sides. Select the standing seam and fold it over then fold in the corners.
Adjust the position of the loops around the top and bottom of the carton. Looking at the references: the top tends to bow inwards and the bottom tends to bow outwards slightly. Loop select the edges that define the perimeter of the folded seam and corners of the box and add support loops with a bevel / chamfer operation. The folded seam on the bottom of the carton is the same as the top, only the flaps are folded inwards instead of outwards.
Though this mesh is water tight, the area under the folded seams will have some potential for overlap and could be further simplified by merging it into the underlying surface. In most cases this shouldn't be required but it is an option for any sort of edge case where the overlap causes issues.
Recap: Gather additional reference information about the folded parts of the object and use that to develop a plan for modeling the shapes. Block out the major forms and resolve the topology flow issues before adding support loops. Try to keep the basic geometry fairly simple, with relatively consistent segment spacing, then let the subdivision do most of the smoothing work around the corners of the folded flaps.
@parikshittiwari
Lots of good advice on how to approach modeling the truncated sphere but when shape accuracy is important it's often beneficial to rely on mathematically consistent primitive geometry as a starting point. There's also nothing inherently wrong with using a sphere that has a central pole. It just requires a bit of adjustment to find the right number of rings and segments to get the geometry to line up with the section planes. Below is an example of what this process and resulting topology could look like.
There's also some situations where it makes more sense to use a regular sphere instead of a quad sphere or icosphere, since it's a lot easier to adjust the number of segments to line up with intersecting geometry. The remainder of partial rings can also be used as support loops. Without having to add additional geometry that would otherwise disturb the segment spacing and quality of the surface.
Quad sphere geometry tends to work well with symmetrical section planes but can run into issues when blending asymmetrical shape intersections. Both topology layouts are viable and deciding which to use really depends on what the adjacent geometry looks like.
@sacboi has a detailed post that explains a different modeling process for cutting holes in quad spheres and also covers why it's important to constrain shape changes to the transitional area between the support loops around the shape intersection.
For your latest shape question: the silhouettes on the reference sheet and the detailed side views seem to show a cut out between the two cylinders. Try using the two, parallel cylinders ahead of that area as a proportional reference then block out the larger cylinder towards the back and see how it looks with the gap that appears between the top and bottom.
I think the answer (as always) is in the realm of "moar geo". If you are modelling an (extreme) closeup of any object with any kind of detailed appearance (especially with sharp, uneven changes, like these sharp folds) you need enough geo to support the attempted level of detail. If you are doing this for like a high quality extreme close up for a commercial, the model you show is very low poly as well as uneven. If it's a game object that will only be seen from a distance, it's fine or a bit too detailed (it could literally be 6 polygons and a texture/normal map). If you want to catch the kind of details shown in the photograph, you need pretty extreme high amounts of geo to hold those subtle deformations. Probably a combination of modelling the basic folds and sculpting the details would be ideal in that case.