For a shape like this, does it matter how many sides your sphere has? I really fail to see any practical relevance to the amount of sides it has.
I don't get what you mean by this question... What do u want to say? Sorry about my English.
Sides? Didn't you try to tell me should go with more sides sphere, because it requires more geo to work well... then you say sphere sides doesn't matter?()... Shape like this? Yes, it's simple one, but at first, I asked, I gave the shape, you said need more sides/segments. Doesn't it relevance anything at all, does it?
Finally, you want to say - In this easy case it doesn't matter how many sides sphere has, then here the method is independent of sphere segments amount?!
Anyway, sorry for my badness in Eng, and thanks again. I'll practise more. U're right, it doesn't take much time to try simple case.
Yes sorry, what I mean to say is *you can't use too many sides*, and if you use enough sides, you can get the appropriate sharpness. Not to say sides do not matter to the model.
Hey guys this discussion intriqued me, so I thought I drop my hat in the ring.
Here is the final
and here is some of the process I used
I hope this look right.
I really started from a box that I divided about
8x8x8
that I put a spherify and then proceed to cut in my shapes and so back and forth on it until I was mostly happy.
From going from second to the final image I collapse most of the stack to work on the topology of cuthing the corners that the shell left to soft. Mostly though had to just move around a few verts as the spherify got me most of the way there but left some things that need to be sorted out.
yes it works, but keep in mind if you smooth a simple quadrangle (square), the result will be not a perfect circle.
If you really need a 'perfect' circle, simply add more geometry to support that shape... this is an example using connect and spherify then extrude in Max (the green circle is a circle spline to test the result):
If you really need a 'perfect' circle, simply add more geometry to support that shape... this is an example using connect and spherify then extrude in Max (the green circle is a circle spline to test the result):
Pedro, I'm using Max... I posted before the same example and it was done way more simple, just a quad. But (as shogunato pointed) TurboSmooth or the other solutions to smooth a quad in Max don't give an absolut 'perfect' circle. So, I needed Connect with '5' segments until get a shape that smooths like a circle spline... a 'perfect' circle. Let me know if you need more pics of what I meant, just try Max and tell me if you find better ways to approach it (please).
With a quad you get a visibly distorted circle. just use a few more sides. not 10 million.
if you want to make an hole into a circle quickly, cap it so it becomes an ngon. select the ngon and run 'geo poly' on it. its in the graphite tools. it's much much better than sphereise.
Not 10 million... just 5 segments to get it match 'perfectly' to a circle spline, a 'perfect' circle. Obviously I don't need a 'perfect' circle, hope you guys see what I was trying to understand with that exercise. And when I was using a quad, I tried to explain that if you don't add more geo you get sphere-like results.
I really started from a box that I divided about
8x8x8
that I put a spherify and then proceed to cut in my shapes and so back and forth on it until I was mostly happy.
From going from second to the final image I collapse most of the stack to work on the topology of cuthing the corners that the shell left to soft. Mostly though had to just move around a few verts as the spherify got me most of the way there but left some things that need to be sorted out.
Thanks Danielmn for spending time to help, it's another way to approach. I'll keep some steps to apply in another model which more complicated. This one I think previous way is fine.
sebas, I think in ur example, quad in quad and subdiv at least level3 to make it smooth. (>10500 faces). Another hand, 8sides in quad just creates <4900 faces in level2 subdiv, looks smooth. For circle in quad, I think method that EQ mentioned is more well know & popular than your 2nd one (that one with many edge connections).
It's appliable for max. Connect vertices make 2 diagonal. Chamfer the center vertex to make 8sides gon.
2 objects in level 2 subdiv. Your quad-in-quad one one need go level3 for smooth but still stretched circle (into a square).
Quick side question, mostly aimed at EarthQuake : I know you use Modo, have you used the Pixar SubD stuff for hard modeling in Modo? Any opinions on it?
nnq2603 thanks for seeing that, yeah putting a circle into a flat object is not the hardest but the method of using a square is the best when it comes to using something flat.
It really depends what the end form needs to be.
While we are on the subject of different methods did you know in max 2011 now has button to make you 8 verts that are in a square .. into a circle. And can be done on curve surface although does not work all the time exactly how you would like its worth giving a try
Quick side question, mostly aimed at EarthQuake : I know you use Modo, have you used the Pixar SubD stuff for hard modeling in Modo? Any opinions on it?
I havent yet, pedro was showing me some images yesterday tho, apparently it just "works", and you dont really have to change how you model or anything, it just produces cleaner results. We'll see if he can post some examples here.
I havent yet, pedro was showing me some images yesterday tho, apparently it just "works", and you dont really have to change how you model or anything, it just produces cleaner results. We'll see if he can post some examples here.
It isnt really portable though is it? I mean any other app will need to be using the same exact subd function
You could freeze the sub-d surface into the mesh to take somewhere else, but yeah, then you lose the ability to edit the base mesh.
Yeah, pretty much. I rarely ever work on my models in a different app, just send them to max/maya/xn to bake normals, so not a huge deal. Tho I'd rather import a 300k cage mesh than a 5million tri raw mesh.
I've a question about sub-d modeling 2 intersection tubes (contains hole, not solid cylinder). (I saw in another page, EQ or someone give tip for solid cylinder but not inside).
A proper/clean way to make it? How. I manage to make it but takes some time to manually slide, cut some faces and weld/collapse vertices (and leave some small inaccurate vertices distance due manual do stuffs). I guess that there's proper & more effective method than mine. The general edge flow and smooth ending result I think OK, but my steps isn't effective enough.
Here is 2 tubes intersection that I want to know your steps:
I've some problems for cleaning inside intersection (where edge intersection/across each other but doesn't created any shared vertex(?)... sorry not sure about words, my bad English. Tried ProBoolean in max & result isn't good). That's where I manually do thing and make out some inaccuracy. I want next time I made it with ease. This case, 2 tubes are same radius.
/Poly count in statis doesn't stand for only this object, something else hidden.
yeah nnq I totaly agree thats the way.
but yes you have to clean that edge up almost every time you use the shell modifier, I wish they would modify to fix that part of the script
another way you could do it is to maybe make a scale down copy, invert the normal move a few edges and use the bridge command to finish it. Either way needs a little clean up effort.
I think I already overlook it... Shell modifier can do that without manually fix afterward. Check "Straighten Corners" and everything should be fine. I search sth about shell in web and found it. I didn't notice there's that option.
I started model a cute robot yesterday. 2D concept by another artist. I haven't got any credit, just personally practise sub-d and want make sth clean, simple. What do u think~ It doesn't exactly look like the concept but a passable form. I need some more tweak...
Hey guys i have been modeling for quite a while so far, nearly like a year so far, and i have never been capable of doing this stupid thing, i know it sounds really nooby but i have been trying to fix it for near a year.
This is the problem, modeling this:
Hey guys i have been modeling for quite a while so far, nearly like a year so far, and i have never been capable of doing this stupid thing, i know it sounds really nooby but i have been trying to fix it for near a year.
This is the problem, modeling this:
Looks like you did alright? Just way too much of an indent, in the ref its barely going in. Probably simpler and cleaner to just make all those indents a floater though
Hey guys, been following these forums/this thread for a while but this is my first post here.
I need a bit of help, I was wondering how you guys would model this tire:
I looks pretty simple at first but it's actually quite complex and I'm having a lot of trouble with it. It's particularly in the middle where the 5 spokes come together and curve down into the bolts (and making those holes for the bolts to sit in).
Can anyone give me some pointers? Or if someones keen enough to make it, make a recording/gif of how you did it?
Hey guys, been following these forums/this thread for a while but this is my first post here.
I need a bit of help, I was wondering how you guys would model this tire:
I looks pretty simple at first but it's actually quite complex and I'm having a lot of trouble with it. It's particularly in the middle where the 5 spokes come together and curve down into the bolts (and making those holes for the bolts to sit in).
Can anyone give me some pointers? Or if someones keen enough to make it, make a recording/gif of how you did it?
Thanks in advance guys, excellent thread.
Firstly don't try the whole thing. Break it up into fifths (72 degrees).
The outside of the rim you could just start with a torus, make it kinda flat, then get the bevels happening.
The center should also be easy; start with a 15-sided cylinder with 2 edge caps, then delete 4/5th of the faces around the center (only 3 left). The venter face of the 3 you have left should be made wider and then extruded inwards so it can accommodate the wheel nut. The center cap is going to be part of the ford badge.
I'm quite new to poly/sub-d modelling (and this forum for that matter) and i'm not really sure how to approach a little project i started the other day.
Basically I want to model a plane from an old Clint Eastwood film called Firefox. The issue I'm having is how to transition from the angular 'chisel' like nose of the aircraft into the rounded fuselage. I can get the basic shape but my edge-loops/bevels are a bit out of control and when I smooth it it loses its definition. There are also a few non-planar surfaces (not just a little) that I can't seem to resolve.
A particular point of focus is under the belt-line near the front of the model there is a triangular shape (which is actually a quad) that im not convinced is the best way to approach.
I would like to get it right now while its pretty simple before i go ahead and increase the resolution etc.
I have attached a few reference images of the plane along with an .obj of what i did the other day. i am using Maya 2011.
scotty
I think this might help
I did a similiar piece really fast that I built out the front part to the way I wanted.
Then counted my sides and made a cylinder with the same amount of sides.
Messed with the cylinder to form the rest of the middle section and mosetly used the soft selction to move to things to keep things smooth.
Then using the bridge brought the two sides together hop this method help.
This is not a good solution. It results in perfectly hard edges, and will not bake well.
I've heard some people say it is, and some people say it isn't a good solution*. It worked in my situation and baked out great. Just trying to help :poly136:
I mean you can do a million different things to get the end result, but doing a proper sds mesh with smart control edges is the best, this is going to be a common problem you'll deal with in sds modeling, its best to simply learn how to deal with it than avoid it.
Exactly. You always see people saying "oh I know this is a crappy solution but it worked for my needs in this particular case". Well, if you learned to do it properly there wouldn't be a need to spend hours hacking your way around it every time.
I never said it was a crappy solution; you guys did. It was a project on a deadline that had to be met and I did what I had to do to meet it with a product I can be proud of.
There are a thousand ways to skin a cat, and next time, when I have more time to do it the "proper way", then I'll be sure to invest the time to do it. Thank you, I always learn so much from the Polycount community.
Lets take a quick look here, this was a very quick test to do, and i'll show a variety of results.
First off, a nice proper sds model with uniform bevels on the edge. This is a simple mesh to create, and again like most issues in this thread, we just need to use enough geometry on the curve to get an accurate result.
Now, comparison to the sds method and the "hard edge" method
Wires on our low test case meshes
Now here we see the results.
A. Excellent, care free bake, no issues
B. Excellent, care free bake, no issues
C. "technically" this method bakes, but we see aliasing and resolution problems, in addition to it just simply not looking very good, this really isn't much better than just using a flat blue normal map on the ridges.
D. Total failure, because we have infinitely sharp angles at 90 degree angles, the sides of the ridges simply do not show up at all. So this method will not work on a more simplified mesh.
Normals, its hard to see the differences between A and C because of the uv layout, but B and D are clear.
You can skin a cat with a rusty butter knife, but its going to be inefficient, and the results aren't going to be pretty.
Hey,
I see that many posts here are about the execution of certain minutes details using subdivision modeling. I personally think it is also very important to be able to roughly establish the main shape in space (using whatever modeling technique fits the need), and then, focus on placing the details on this surface. Also having the big "parent" shape available for reference somewhere in the scene allows for very easy "just add more geo" passes.
Here is an exemple using mudbox, topogun and max. I know there is some pinching left, but you get the idea. Here I am mostly focusing on speed of execution anyways (and that is an important aspect of game art production!)
This is obviously not that useful for shapes described in detail in ortho views, like cars and such. But whenever you have to work from a somehow organic piece of concept art presented in perspective, using a sculpting program to establish the main forms is a life saver!
I'd be interested to see how you get it to match the grooves in the reference. You can see the top edge (and all the edges, really) is very sharp, which is why this problem was so tricky. Keeping those edges sharp, while the long bevel stayed soft proved very troublesome.
If I were to use soft edges like you did, I don't think it would've been an issue at all.
Again, I appreciate your time, as you have gotten closer to anyone else I've seen throw examples out there. I'll be sure to use better geometry from the start next time.
Simply raise a few edges and we can tighten that edge with the same toplogoly, its very important however to remember 2 things:
1. Zoom out! Analyzing edge hardness when zoomed way into your model is always a bad idea. While it may look "soft" zoomed in here, pay attention to the last frame, viewed from a reasonable distance, it is plenty sharp.
2. You're limited to texture resolution as far as how sharp a line can be so its best to make models that are "soft" enough to actually show up in the bake, otherwise you get aliasing. Its always more important to have an asset that bakes well and reads well as apposed to having an "accurate" highpoly that is messy in the end result.
I've had a lot of realism gun nerds try to argue with me on this point, but they all eventually figure it out for themselves the when they get more experience. Technical limitations trump anal retentive desires for accuracy.
Now, if you still want it harder, its just a matter of using even more sides on your curved surface here to add in more natural supporting edges that A. define the curve but also B. add sharpness where you want it. So, just a matter of how you specifically implement it. Tweaking how tight your supporting edgers here, in general, will also work to harden various elements without being destructive.
But again, because of the specifics of texture resolution, this may be a waste of your time.
If we look at what is actually happening in the normal, we'll see that this angle change is basically represented with 1 column of pixels. So, you literally can not get any sharper without increasing the resolution of the image.
Just another quick image, vary your cylinder resolution and supporting edges = varied edge hardness. This is a basis SDS principal that will be applicable in a wide variety of situations. Again just matter of how you specifically decide to do it.
Started from 48, 36, 24 and 16 sides, ignore mr. 36, he should have been 32.
That makes much more sense to me now. I see now that I was obsessing over something that wasn't even going to matter in the end. Oh well, smh. I'm still learning so forgive me :poly136: Thanks for your help and patience.
Replies
I don't get what you mean by this question... What do u want to say? Sorry about my English.
Sides? Didn't you try to tell me should go with more sides sphere, because it requires more geo to work well... then you say sphere sides doesn't matter?()... Shape like this? Yes, it's simple one, but at first, I asked, I gave the shape, you said need more sides/segments. Doesn't it relevance anything at all, does it?
Finally, you want to say - In this easy case it doesn't matter how many sides sphere has, then here the method is independent of sphere segments amount?!
Anyway, sorry for my badness in Eng, and thanks again. I'll practise more. U're right, it doesn't take much time to try simple case.
Here is the final
and here is some of the process I used
I hope this look right.
I really started from a box that I divided about
8x8x8
that I put a spherify and then proceed to cut in my shapes and so back and forth on it until I was mostly happy.
From going from second to the final image I collapse most of the stack to work on the topology of cuthing the corners that the shell left to soft. Mostly though had to just move around a few verts as the spherify got me most of the way there but left some things that need to be sorted out.
Yes, 3x more sides precisely. With 3x more you get natural supporting edges along the shape as well.
If you really need a 'perfect' circle, simply add more geometry to support that shape... this is an example using connect and spherify then extrude in Max (the green circle is a circle spline to test the result):
Damn, you guys sure like to complicate.
Pedro, I'm using Max... I posted before the same example and it was done way more simple, just a quad. But (as shogunato pointed) TurboSmooth or the other solutions to smooth a quad in Max don't give an absolut 'perfect' circle. So, I needed Connect with '5' segments until get a shape that smooths like a circle spline... a 'perfect' circle. Let me know if you need more pics of what I meant, just try Max and tell me if you find better ways to approach it (please).
if you want to make an hole into a circle quickly, cap it so it becomes an ngon. select the ngon and run 'geo poly' on it. its in the graphite tools. it's much much better than sphereise.
4 a.m. here... going to bed.
sebas, I think in ur example, quad in quad and subdiv at least level3 to make it smooth. (>10500 faces). Another hand, 8sides in quad just creates <4900 faces in level2 subdiv, looks smooth. For circle in quad, I think method that EQ mentioned is more well know & popular than your 2nd one (that one with many edge connections).
It's appliable for max. Connect vertices make 2 diagonal. Chamfer the center vertex to make 8sides gon.
2 objects in level 2 subdiv. Your quad-in-quad one one need go level3 for smooth but still stretched circle (into a square).
It really depends what the end form needs to be.
While we are on the subject of different methods did you know in max 2011 now has button to make you 8 verts that are in a square .. into a circle. And can be done on curve surface although does not work all the time exactly how you would like its worth giving a try
here is something else that tool and do
I havent yet, pedro was showing me some images yesterday tho, apparently it just "works", and you dont really have to change how you model or anything, it just produces cleaner results. We'll see if he can post some examples here.
It isnt really portable though is it? I mean any other app will need to be using the same exact subd function
You could freeze the sub-d surface into the mesh to take somewhere else, but yeah, then you lose the ability to edit the base mesh.
ps: Shouldn't meshsmooth work the same in all applications? Multiplying by a power of 2 and averaging distances based on existing vertices...
Yeah, pretty much. I rarely ever work on my models in a different app, just send them to max/maya/xn to bake normals, so not a huge deal. Tho I'd rather import a 300k cage mesh than a 5million tri raw mesh.
A proper/clean way to make it? How. I manage to make it but takes some time to manually slide, cut some faces and weld/collapse vertices (and leave some small inaccurate vertices distance due manual do stuffs). I guess that there's proper & more effective method than mine. The general edge flow and smooth ending result I think OK, but my steps isn't effective enough.
Here is 2 tubes intersection that I want to know your steps:
I've some problems for cleaning inside intersection (where edge intersection/across each other but doesn't created any shared vertex(?)... sorry not sure about words, my bad English. Tried ProBoolean in max & result isn't good). That's where I manually do thing and make out some inaccuracy. I want next time I made it with ease. This case, 2 tubes are same radius.
/Poly count in statis doesn't stand for only this object, something else hidden.
Make tubes connect outside. And remove all circle faces, then apply Shell.
But I notice Shell make out unparallel edges. Then have fix to it with MakePlanar for selected vertices.
but yes you have to clean that edge up almost every time you use the shell modifier, I wish they would modify to fix that part of the script
another way you could do it is to maybe make a scale down copy, invert the normal move a few edges and use the bridge command to finish it. Either way needs a little clean up effort.
I started model a cute robot yesterday. 2D concept by another artist. I haven't got any credit, just personally practise sub-d and want make sth clean, simple. What do u think~ It doesn't exactly look like the concept but a passable form. I need some more tweak...
cant believe I missed that one. Has that been there for a while or is it new to 2011 either way thanks man.
keep up the good work
This is the problem, modeling this:
Looks like you did alright? Just way too much of an indent, in the ref its barely going in. Probably simpler and cleaner to just make all those indents a floater though
the proportions might be a little off. Let me know if you need anything else
I need a bit of help, I was wondering how you guys would model this tire:
I looks pretty simple at first but it's actually quite complex and I'm having a lot of trouble with it. It's particularly in the middle where the 5 spokes come together and curve down into the bolts (and making those holes for the bolts to sit in).
Can anyone give me some pointers? Or if someones keen enough to make it, make a recording/gif of how you did it?
Thanks in advance guys, excellent thread.
Thanks for your time and help
you would probably need to start from a circle of around 12 sides just for a fifth of the model.
So 60 for the entire circle at least
Firstly don't try the whole thing. Break it up into fifths (72 degrees).
The outside of the rim you could just start with a torus, make it kinda flat, then get the bevels happening.
The center should also be easy; start with a 15-sided cylinder with 2 edge caps, then delete 4/5th of the faces around the center (only 3 left). The venter face of the 3 you have left should be made wider and then extruded inwards so it can accommodate the wheel nut. The center cap is going to be part of the ford badge.
I'm quite new to poly/sub-d modelling (and this forum for that matter) and i'm not really sure how to approach a little project i started the other day.
Basically I want to model a plane from an old Clint Eastwood film called Firefox. The issue I'm having is how to transition from the angular 'chisel' like nose of the aircraft into the rounded fuselage. I can get the basic shape but my edge-loops/bevels are a bit out of control and when I smooth it it loses its definition. There are also a few non-planar surfaces (not just a little) that I can't seem to resolve.
A particular point of focus is under the belt-line near the front of the model there is a triangular shape (which is actually a quad) that im not convinced is the best way to approach.
I would like to get it right now while its pretty simple before i go ahead and increase the resolution etc.
I have attached a few reference images of the plane along with an .obj of what i did the other day. i am using Maya 2011.
Thanks in advance for any help offered.
scotty_b
http://dl.dropbox.com/u/8296143/Firefox_mesh_b.obj
[img][/img]
I think this might help
I did a similiar piece really fast that I built out the front part to the way I wanted.
Then counted my sides and made a cylinder with the same amount of sides.
Messed with the cylinder to form the rest of the middle section and mosetly used the soft selction to move to things to keep things smooth.
Then using the bridge brought the two sides together hop this method help.
This is not a good solution. It results in perfectly hard edges, and will not bake well.
I've heard some people say it is, and some people say it isn't a good solution*. It worked in my situation and baked out great. Just trying to help :poly136:
*"The Debate" On GameArtisans
I never said it was a crappy solution; you guys did. It was a project on a deadline that had to be met and I did what I had to do to meet it with a product I can be proud of.
There are a thousand ways to skin a cat, and next time, when I have more time to do it the "proper way", then I'll be sure to invest the time to do it. Thank you, I always learn so much from the Polycount community.
First off, a nice proper sds model with uniform bevels on the edge. This is a simple mesh to create, and again like most issues in this thread, we just need to use enough geometry on the curve to get an accurate result.
Link to obj: http://dl.dropbox.com/u/499159/glocktestcage.obj
Now, comparison to the sds method and the "hard edge" method
Wires on our low test case meshes
Now here we see the results.
A. Excellent, care free bake, no issues
B. Excellent, care free bake, no issues
C. "technically" this method bakes, but we see aliasing and resolution problems, in addition to it just simply not looking very good, this really isn't much better than just using a flat blue normal map on the ridges.
D. Total failure, because we have infinitely sharp angles at 90 degree angles, the sides of the ridges simply do not show up at all. So this method will not work on a more simplified mesh.
Normals, its hard to see the differences between A and C because of the uv layout, but B and D are clear.
You can skin a cat with a rusty butter knife, but its going to be inefficient, and the results aren't going to be pretty.
I see that many posts here are about the execution of certain minutes details using subdivision modeling. I personally think it is also very important to be able to roughly establish the main shape in space (using whatever modeling technique fits the need), and then, focus on placing the details on this surface. Also having the big "parent" shape available for reference somewhere in the scene allows for very easy "just add more geo" passes.
Here is an exemple using mudbox, topogun and max. I know there is some pinching left, but you get the idea. Here I am mostly focusing on speed of execution anyways (and that is an important aspect of game art production!)
http://img.photobucket.com/albums/v242/pior_ubb/mecha_001_wip001.png
http://img.photobucket.com/albums/v242/pior_ubb/mecha_001_wip002.png
http://img.photobucket.com/albums/v242/pior_ubb/mecha_001_wip003.png
This is obviously not that useful for shapes described in detail in ortho views, like cars and such. But whenever you have to work from a somehow organic piece of concept art presented in perspective, using a sculpting program to establish the main forms is a life saver!
I'd be interested to see how you get it to match the grooves in the reference. You can see the top edge (and all the edges, really) is very sharp, which is why this problem was so tricky. Keeping those edges sharp, while the long bevel stayed soft proved very troublesome.
If I were to use soft edges like you did, I don't think it would've been an issue at all.
Again, I appreciate your time, as you have gotten closer to anyone else I've seen throw examples out there. I'll be sure to use better geometry from the start next time.
1. Zoom out! Analyzing edge hardness when zoomed way into your model is always a bad idea. While it may look "soft" zoomed in here, pay attention to the last frame, viewed from a reasonable distance, it is plenty sharp.
2. You're limited to texture resolution as far as how sharp a line can be so its best to make models that are "soft" enough to actually show up in the bake, otherwise you get aliasing. Its always more important to have an asset that bakes well and reads well as apposed to having an "accurate" highpoly that is messy in the end result.
I've had a lot of realism gun nerds try to argue with me on this point, but they all eventually figure it out for themselves the when they get more experience. Technical limitations trump anal retentive desires for accuracy.
Now, if you still want it harder, its just a matter of using even more sides on your curved surface here to add in more natural supporting edges that A. define the curve but also B. add sharpness where you want it. So, just a matter of how you specifically implement it. Tweaking how tight your supporting edgers here, in general, will also work to harden various elements without being destructive.
But again, because of the specifics of texture resolution, this may be a waste of your time.
If we look at what is actually happening in the normal, we'll see that this angle change is basically represented with 1 column of pixels. So, you literally can not get any sharper without increasing the resolution of the image.
Started from 48, 36, 24 and 16 sides, ignore mr. 36, he should have been 32.