@Fjr3 Isn't the front a flat glass plane with a hole cut in it? Why not model it like a flat plane with a hole in it? If two things are seperate objects, why not model them that way? As for why it's fucky, your problem is twofold: Your cylinder isn't shaped like a cylinder and it's sides are uneven, which is why it smooths like an egg. You also don't have proper supportive edgeloops around it to force the shape to stop, so it gets that sideways pinchy effect. This hole has eight equal sides, and has supportive edgeloops to constrict the subdivision so it'll smooth properly.
nail the shape first, subdivide for more res, do the holes. once those are locked in, you should not have to touch the shape
Thanks for the answer. I will work on that.
I also had issues with another form on this personal project. The cover for the particle filter has spokes that goes towards the middle. The spokes are welded together with circular forms/insets at regular intervals. On the bigger filter I started with the rim of the filter and extruded faces towards the middle to create the spokes. The rim has a circular shape but the rim "inset" has straight lines. I want to insets to be circular just like on the reference photo. This is a very common shape but somehow I just got stuck :P thanks in advance.
How do I slice models like this video at 16:05 in 3ds max? He is not using the Slice Plane tool, or quick slice, at least I think he isn't. He is basically just slicing a whole object on an axis much like the Quick Slice tool.
Going for that front lights part. At first I wanted to use more sided cylinder and not subdivisions, as rest of the model doesnt need subdividing. And as far as I know, every face should be 4-quad or triangle. Cant get it here, so I wonder whats the best way to do this part.
Going for that front lights part. At first I wanted to use more sided cylinder and not subdivisions, as rest of the model doesnt need subdividing. And as far as I know, every face should be 4-quad or triangle. Cant get it here, so I wonder whats the best way to do this part.
Filip, build the main body. (the bigger parts). Move on to the secondary part of the car (the second biggest shape). Use that as a cage before cutting holes .
Insert a loop on each of those flat rims, scale outwards until roundness is achieved.
Thanks for the answer. I appreciate your answer but your solution means that I should eyeball the roundness? I tried a different technique today on the small air exhaust on the gasmask. I started with a sphere. Then I selected a couple of edgeloops on the sphere. Then I did a quad chamfer on the selected lines. Then I extruded the chamfered lines. With the quad chamfer the top row had an equal distance between the segments.
Yeah why not? Use a reference cylinder and match it if you want. Whatever gets you there.
As for that protective cap, you can totes use a sphere for it, but squash it first as it's not spherical... more like a zeppelin shape. Get good refs, study them and follow them. Nail the large shapes first, then do the details. Make sure you have enough geometry to help out the cuts you're about to do.
I hope you have max 2017, because I did this thing but would rather just let you browse through the file than do a bunch of pics in mspaint. 15 steps to victory: http://skins.thanez.net/help/johan_filter.max
Notes: There are two main objects because I had to collapse the stack because longstack-max files like to crash on this laptop for some reason. There's a third object called doublesmooth method. It's a copout alternative to step 15. IDK use it if you want. Both ways create unique minor errors that won't be seen on a matte plastic material. About step 15, it should really happen right after step 7, but I fucked up and couldn't be arsed to go back. Bad planning right there led to me having to split off the tilable pattern, delete all but 1, do the corners of the cutout, repeat the pattern and weld it back on. Step 0: The cylinder side amount is a 6x multiplier because you have a six-sided circular pattern to model. I chose 6x8 because that way the bits that go from the center to the extremities gets a center edge that gives it roundness. Step 4: If you doubled this, the circular ribs would be rounded, and would have their own supportive geometry, so you could skip step 6 and 15. In retrospect this would be optimal.
Yeah why not? Use a reference cylinder and match it if you want. Whatever gets you there.
As for that protective cap, you can totes use a sphere for it, but squash it first as it's not spherical... more like a zeppelin shape. Get good refs, study them and follow them. Nail the large shapes first, then do the details. Make sure you have enough geometry to help out the cuts you're about to do.
I hope you have max 2017, because I did this thing but would rather just let you browse through the file than do a bunch of pics in mspaint.
Thanks a lot! number 2. the set flow step really makes sense. That makes sure you have an even distribution of the edgeloops after the squash. On step 9. array to rebuild pattern I would use the clone modifier( I am sure you are using that one as well). The problem I have with your solution is the bits that go from the center to extremities. These bits are getting smaller towards the center in your model. This is something I wanted to avoid as they look straight in the reference. This is even more visible in the large particle filter cover.
after step 4, select those edges, constrain by edge, make planar by axis.
Instructions unclear, fingers stuck in toaster ... just joking
thanks, I definitely noticed a difference when I made them planar by axis. I left out the last row to keep the top part intact. just a quick test ( in reality I only need to make one edge planar due to the symmetry modifier etc.). It was pretty cool to notice that I can change the squish and it propagates all the way through the modifier stack. When it comes to reference I have a pretty decent camera and the plan is to take photos of all the equipment I am modelling. I guess it would speed up the process to have a lot of refs on the same object. To get a correct feeling of the form.
its not perfect. but. 1. Main shape, sphere wit 18 sides. 2. turbo smoothed once 3. Added where the cuts would be (mostly deleting faces) and always working on 1/4 or 1/7 of the mesh 4. Rotated the copies and welded vertices. 5 minutes of work.
Double the amount of segments in the cylinder and you should be fine with that approach. I use it myself a lot of the time, trying to avoid any hand-modeled control edges whatsoever. The resulting artifacts are typically too small to be noticed.
I want to challenge you on your use of Edit Poly mods though. They are not parametric, so why leave them on the stack? That looks very questionable and likely to only lead to problems.
Hi all, I'm a relatively novice modeller and some time ago I thought I'd try and tackle modelling a WWII British Matilda II tank. I've made some decent progress but some shapes in the turret are giving me a headache. I am now struggling with the front of the turret ring/base which is a really particular shape. I've attached a couple of reference images and my progress so far. This actually looks allright and might be just enough geo to look good from a distance, but I'm wondering: A: What is a good approach to model a piece like this? (it took me a lot of time, using booleans to give me the basic shape, then moving a lot of points around manually. But perhaps this is already the best way) B: if I wanted to subdivide this, what would be my edgeflow around the highlighted area?
I want to challenge you on your use of Edit Poly mods though. They are not parametric, so why leave them on the stack? That looks very questionable and likely to only lead to problems.
Hey perna, can you expand on this a little? I'm semi-new to subd and am wondering. Do you just work on 1 editable poly, or collapse your stack as needed? How do you keep it parametric?
Double the amount of segments in the cylinder and you should be fine with that approach. I use it myself a lot of the time, trying to avoid any hand-modeled control edges whatsoever. The resulting artifacts are typically too small to be noticed.
I want to challenge you on your use of Edit Poly mods though. They are not parametric, so why leave them on the stack? That looks very questionable and likely to only lead to problems.
I doubled the amount of segments but I still notice some small weirdness when I zoom in see something like this.
@snakedogman Firstly that's a tough subject. It doesn't have clean, machined shapes. I would lose hair working on something like that; you might want to try something cleaner. I've re-created your shape (not ref) with simple sub-d modeling in 3ds max using a subdivision modifier set to respect smoothing groups.
A - cylinder primitive, 32 sides as that lines up nicely with your ref shapes (orange) B - inset then extrude on the bottom ngon C - two manual cuts (marked yellow in the second column), deletion of polygons D - fill the gap (shift-drag edges, target weld)
So there's barely any manual work and consequently it's quick and the result is accurate.
@Turks It might sound like a joke answer, but I leave modifiers on the stack when it's beneficial to do so. So I'm asking the above person what the benefit is of his use of "Edit Poly" modifiers.
You could spin around and rub your belly every time you take your shoes off because someone on the internet told you that's a better way of removing shoes. But you can't do stuff just because someone (like me) says you should. You need to understand exactly why something might be beneficial to do. It's like all these people who model purely with quads. They never stop to think why, they just blindly follow some random stuff they read online. Now let's talk about religion.
@johanflod But now you're no longer using the technique you outlined in your image.
Here are some variations. The visual artifacts are small. You're going to get artifacts like that with any normal map bake anyway. Unless you're making a huge chrome shape you can get away with it. Never aim to make subd-meshes perfect, just aim to make them good enough for their intended purpose, otherwise you're going to waste a lot of time.
See how messy the last mesh is, with those huge chamfers. Works just fine still.
As for whether the new stack is better. Well, why ask me, I'm not the one working with it. Is it better for you?
@snakedogman Firstly that's a tough subject. It doesn't have clean, machined shapes. I would lose hair working on something like that; you might want to try something cleaner.
You're absolutely right Perna. Of course, being a novice, I didn't realise when I started this model, it would be quite so hard. I had started working on the body of the tank which was ok, but the turret is rather a challenge, since it's this weird, cast iron shape. Then again, if we don't set ourselves some challenges, how do we grow eh? But thanks a lot for the example, I'm gonna study it!
@Turks It might sound like a joke answer, but I leave modifiers on the stack when it's beneficial to do so. So I'm asking the above person what the benefit is of his use of "Edit Poly" modifiers.
You could spin around and rub your belly every time you take your shoes off because someone on the internet told you that's a better way of removing shoes. But you can't do stuff just because someone (like me) says you should. You need to understand exactly why something might be beneficial to do. It's like all these people who model purely with quads. They never stop to think why, they just blindly follow some random stuff they read online. Now let's talk about religion.
haha that makes sense. That's a practice I've tried to follow when making anything in any medium. Just was wondering if you had any specific wisdom to impart particularly about the parametric aspect. Didn't mean to set you off, jk
@johanflod But now you're no longer using the technique you outlined in your image.
Here are some variations. The visual artifacts are small. You're going to get artifacts like that with any normal map bake anyway. Unless you're making a huge chrome shape you can get away with it. Never aim to make subd-meshes perfect, just aim to make them good enough for their intended purpose, otherwise you're going to waste a lot of time.
See how messy the last mesh is, with those huge chamfers. Works just fine still.
As for whether the new stack is better. Well, why ask me, I'm not the one working with it. Is it better for you?
There are a few modifiers I would keep in the stack like turbosmooth, chamfer, shell and clone. Having two edit poly modifiers on top of each other does not make sense and I would merge them. Leaving the cylinder modifier at the bottom is like asking for trouble. You change the amount of sides on the cylinder and you break everything above. I added the turbo smooth modifier ( with separate by smoothing groups checked ) before the chamfer to quickly test if more subdivisions would make the form hold up better. Why am I no longer using the technique outlined in the image?
@johanflod Hi, try this as mine.... Its without any pinch, and I am sure there will be other ways.
Yes to get it perfect you need to do as you did.
Could either of you guys show how that's supposed to be perfect? Also, what happened to using smoothing groups?
Perfect is the wrong word from my side and I changed it. Communicating over the intrawebs demands the right words at all places I meant that @iacdxb did not have any pinching in his solution. I would still prefer a non destructive method. I did an experiment with the standard chamfer(instead of the quad option in the modifier) and smoothing groups.
@johanflod here are both your topologies. The results are almost identical. If you're getting pinching it's not because of the topology but because you're keeping control loops unnecessarily close together in the areas marked in purple.
Be wary of spending time chasing the perfect solution with subd. Also keep in mind that with the chamfer modifier, since you're not making control loops manually (and spending ungodly amounts of time doing it), it's often irrelevant how dense your base geometry is, so you could in this case use 64 or however many segments for the cylinder.
That's what I figured as well. It could also conceivably be helpful to have such regularized meshes when it comes to texturing and shading with their internal tools, but I'm taking intuitive pot-shots.
Damn if they aren't some beautiful examples of quadrangulated meshes though.
Hey there, saw this shape and tried doing it. Posting here to see if I'm going in the right direction or if the answer is easier somehow.
-So I make the base shape (colored object) with booleans and edge bevel -Manual tweak by deleting edge (white arrow) and smooth using chamfer+SGs.
Trying to be nitpicky, I say that looking at it from the top , the bevel width isn't consistent, as it thins around the circular edge. I know I could get acceptable results by manually tweaking that too, but I'm starting to think there must be a more mechanical approach to this piece...
Per would maybe suggest to use the Catmull-Clark principle of subdivision to mathematically get the roundness right, but I struggle to figure out how to do that without needing to eyeball verts.
-Exactly why did you choose that intersection point between the box and cylinder shapes? -Exactly why did you remove those white arrow edges? -Exactly why did you use booleans?
Whether or not you have since reconsidered your approach, it would be great if you could answer those questions as to your original motivations.
Sculpting is not specific to film and you may well make the argument that they try to avoid dense meshes more than we do.
There's no mystery here. Film and games are the same; both use final rendered meshes which are very clean, with an orderly distribution of geometry so they can be put through animation and fx.
-I'm worried I'm not getting you with this question. I picked an 8 sided cylinder for the boolean operation because I thought the number of sides aligned the best with my borders. -I actually first did the shape with those white arrow edges. Just before I was about to post it here, and answering your question, I thought to myself the subd cage would be better without it, since less is more flexible/easier to work with. I also tried making the shape with two edges for the circular sides, but thought 3 would be the best balance. -For the beginning shape? I thought it the fastest way to get me the object I wanted to smooth. Is there a faster approach you can think of?
"he works at ILM" is not enough but if there's some special circumstance in that process that must eliminate ngons I'd like to know. To me if you sub-d any ngon twice you get quads anyway.
Once Read my reply. We don't use ngons, so there's no reason why film should. You want to avoid shading errors, unresolved UVs, animation issues, etc.
@Justo , just to clear up the first question: The intersection point is where the two boolean operators line up. So the question is - why did you decide to make the subtraction exactly there, as opposed to a bit to the left or right or whatever.
Sculpting is not specific to film and you may well make the argument that they try to avoid dense meshes more than we do.
There's no mystery here. Film and games are the same; both use final rendered meshes which are very clean, with an orderly distribution of geometry so they can be put through animation and fx.
Right, I was just meaning that the clean meshes are useful for sculpting since they can be passed over to the animation teams easily (and back and forth between teams) without needing further retopo.
@TeriyakiStyle Ptex works just fine with ngons and triangles, which I checked after making that post suggesting possible texture benefits. It would actually be fine, other than this next point.
It is definitely better to have mostly quads for consistent displacement with subdivision at render time. High end renderers will often subdivide adaptively based on pixels, and having too irregular meshes with strong displacement maps will result in visible jumps in displacement quality. This can also impact how quickly and cleanly the render converges.
Replies
If two things are seperate objects, why not model them that way?
As for why it's fucky, your problem is twofold: Your cylinder isn't shaped like a cylinder and it's sides are uneven, which is why it smooths like an egg.
You also don't have proper supportive edgeloops around it to force the shape to stop, so it gets that sideways pinchy effect.
This hole has eight equal sides, and has supportive edgeloops to constrict the subdivision so it'll smooth properly.
I also had issues with another form on this personal project. The cover for the particle filter has spokes that goes towards the middle. The spokes are welded together with circular forms/insets at regular intervals. On the bigger filter I started with the rim of the filter and extruded faces towards the middle to create the spokes. The rim has a circular shape but the rim "inset" has straight lines. I want to insets to be circular just like on the reference photo. This is a very common shape but somehow I just got stuck :P thanks in advance.
He is not using the Slice Plane tool, or quick slice, at least I think he isn't. He is basically just slicing a whole object on an axis much like the Quick Slice tool.
Going for that front lights part. At first I wanted to use more sided cylinder and not subdivisions, as rest of the model doesnt need subdividing. And as far as I know, every face should be 4-quad or triangle. Cant get it here, so I wonder whats the best way to do this part.
quoting Neox
Yeah why not? Use a reference cylinder and match it if you want. Whatever gets you there.
As for that protective cap, you can totes use a sphere for it, but squash it first as it's not spherical... more like a zeppelin shape. Get good refs, study them and follow them. Nail the large shapes first, then do the details. Make sure you have enough geometry to help out the cuts you're about to do.
I hope you have max 2017, because I did this thing but would rather just let you browse through the file than do a bunch of pics in mspaint.
15 steps to victory: http://skins.thanez.net/help/johan_filter.max
Notes:
There are two main objects because I had to collapse the stack because longstack-max files like to crash on this laptop for some reason.
There's a third object called doublesmooth method. It's a copout alternative to step 15. IDK use it if you want. Both ways create unique minor errors that won't be seen on a matte plastic material.
About step 15, it should really happen right after step 7, but I fucked up and couldn't be arsed to go back. Bad planning right there led to me having to split off the tilable pattern, delete all but 1, do the corners of the cutout, repeat the pattern and weld it back on.
Step 0: The cylinder side amount is a 6x multiplier because you have a six-sided circular pattern to model. I chose 6x8 because that way the bits that go from the center to the extremities gets a center edge that gives it roundness.
Step 4: If you doubled this, the circular ribs would be rounded, and would have their own supportive geometry, so you could skip step 6 and 15. In retrospect this would be optimal.
thanks, I definitely noticed a difference when I made them planar by axis. I left out the last row to keep the top part intact. just a quick test ( in reality I only need to make one edge planar due to the symmetry modifier etc.). It was pretty cool to notice that I can change the squish and it propagates all the way through the modifier stack.
When it comes to reference I have a pretty decent camera and the plan is to take photos of all the equipment I am modelling. I guess it would speed up the process to have a lot of refs on the same object. To get a correct feeling of the form.
I had to include a screenshot of the scene I got from Thanez . Those modifier names!
1. Main shape, sphere wit 18 sides.
2. turbo smoothed once
3. Added where the cuts would be (mostly deleting faces) and always working on 1/4 or 1/7 of the mesh
4. Rotated the copies and welded vertices. 5 minutes of work.
24 subdivide cylinder extrude for nozzle, I know its wrong subdivide from nozzle edge. some triangles but if anyone guide me... much appreciated.
Thanks.
Thanks.
...
To define a hard edge, simply place control edges on either side of it. This is fundamental to sub-d modeling.
...
Double the amount of segments in the cylinder and you should be fine with that approach. I use it myself a lot of the time, trying to avoid any hand-modeled control edges whatsoever. The resulting artifacts are typically too small to be noticed.
I want to challenge you on your use of Edit Poly mods though. They are not parametric, so why leave them on the stack? That looks very questionable and likely to only lead to problems.
A: What is a good approach to model a piece like this? (it took me a lot of time, using booleans to give me the basic shape, then moving a lot of points around manually. But perhaps this is already the best way)
B: if I wanted to subdivide this, what would be my edgeflow around the highlighted area?
(I know my current attempt doesn't have quite the right proportions yet, but I'm confident that if I get the approach right I'll be able to get it closer to the reference)
https://www.dropbox.com/s/ee28i1izxufwfei/matilda%20turret%20blocking%2008b.obj?dl=0
Hey perna, can you expand on this a little? I'm semi-new to subd and am wondering. Do you just work on 1 editable poly, or collapse your stack as needed? How do you keep it parametric?
when I zoom in see something like this.
collapsed part of the stack. is this better?
Firstly that's a tough subject. It doesn't have clean, machined shapes. I would lose hair working on something like that; you might want to try something cleaner. I've re-created your shape (not ref) with simple sub-d modeling in 3ds max using a subdivision modifier set to respect smoothing groups.
A - cylinder primitive, 32 sides as that lines up nicely with your ref shapes (orange)
B - inset then extrude on the bottom ngon
C - two manual cuts (marked yellow in the second column), deletion of polygons
D - fill the gap (shift-drag edges, target weld)
So there's barely any manual work and consequently it's quick and the result is accurate.
@Turks
It might sound like a joke answer, but I leave modifiers on the stack when it's beneficial to do so. So I'm asking the above person what the benefit is of his use of "Edit Poly" modifiers.
You could spin around and rub your belly every time you take your shoes off because someone on the internet told you that's a better way of removing shoes. But you can't do stuff just because someone (like me) says you should. You need to understand exactly why something might be beneficial to do. It's like all these people who model purely with quads. They never stop to think why, they just blindly follow some random stuff they read online. Now let's talk about religion.
But now you're no longer using the technique you outlined in your image.
Here are some variations. The visual artifacts are small. You're going to get artifacts like that with any normal map bake anyway. Unless you're making a huge chrome shape you can get away with it. Never aim to make subd-meshes perfect, just aim to make them good enough for their intended purpose, otherwise you're going to waste a lot of time.
See how messy the last mesh is, with those huge chamfers. Works just fine still.
As for whether the new stack is better. Well, why ask me, I'm not the one working with it. Is it better for you?
Hi, try this as mine.... Its without any pinch, and I am sure there will be other ways.
I did an experiment with the standard chamfer(instead of the quad option in the modifier) and smoothing groups.
here are both your topologies. The results are almost identical. If you're getting pinching it's not because of the topology but because you're keeping control loops unnecessarily close together in the areas marked in purple.
Be wary of spending time chasing the perfect solution with subd. Also keep in mind that with the chamfer modifier, since you're not making control loops manually (and spending ungodly amounts of time doing it), it's often irrelevant how dense your base geometry is, so you could in this case use 64 or however many segments for the cylinder.
Damn if they aren't some beautiful examples of quadrangulated meshes though.
-So I make the base shape (colored object) with booleans and edge bevel
-Manual tweak by deleting edge (white arrow) and smooth using chamfer+SGs.
Trying to be nitpicky, I say that looking at it from the top , the bevel width isn't consistent, as it thins around the circular edge. I know I could get acceptable results by manually tweaking that too, but I'm starting to think there must be a more mechanical approach to this piece...
Per would maybe suggest to use the Catmull-Clark principle of subdivision to mathematically get the roundness right, but I struggle to figure out how to do that without needing to eyeball verts.
-Exactly why did you choose that intersection point between the box and cylinder shapes?
-Exactly why did you remove those white arrow edges?
-Exactly why did you use booleans?
Whether or not you have since reconsidered your approach, it would be great if you could answer those questions as to your original motivations.
sculpting is a valid point
There's no mystery here. Film and games are the same; both use final rendered meshes which are very clean, with an orderly distribution of geometry so they can be put through animation and fx.
-I'm worried I'm not getting you with this question. I picked an 8 sided cylinder for the boolean operation because I thought the number of sides aligned the best with my borders.
-I actually first did the shape with those white arrow edges. Just before I was about to post it here, and answering your question, I thought to myself the subd cage would be better without it, since less is more flexible/easier to work with. I also tried making the shape with two edges for the circular sides, but thought 3 would be the best balance.
-For the beginning shape? I thought it the fastest way to get me the object I wanted to smooth. Is there a faster approach you can think of?
Thanks for your time in teaching us all man.
@TeriyakiStyle Ptex works just fine with ngons and triangles, which I checked after making that post suggesting possible texture benefits. It would actually be fine, other than this next point.
It is definitely better to have mostly quads for consistent displacement with subdivision at render time. High end renderers will often subdivide adaptively based on pixels, and having too irregular meshes with strong displacement maps will result in visible jumps in displacement quality. This can also impact how quickly and cleanly the render converges.