Hello Guys, little turbosmooth question that has been bugging me for a while :
Got these square extrude of a cylinder, did a chamfer. Would like to turbosmooth but I have some really weird thing happening (screen). If I add control loop near those edges, its end up (logically) having a hightlight, which I dont want.
How do I manage that please? Do I have to separe these pieces of the cylinder?
Hello Guys, little turbosmooth question that has been bugging me for a while :
Got these square extrude of a cylinder, did a chamfer. Would like to turbosmooth but I have some really weird thing happening (screen). If I add control loop near those edges, its end up (logically) having a hightlight, which I dont want.
How do I manage that please? Do I have to separe these pieces of the cylinder?
is it really built like that? One mesh for that thing, or is it acctually seperate pieces in real life?
Hello @MaxHoek Thanks for the tips ! I was already using double smooth though, but lacked the supportive edge at right spot, now its better.
Because of the extrude my edge flow wasn't really good and I had to do a new mesh with highter geo before extruding those part. Then I did double smooth and it worked way better, although I Still have some pinching... maybe i'm still missing some geo?
@perna Hello Perna, I'm sorry you guys have to answers again and again the same questions, I didn't knew that my cases could be issues with Wirrexx example. Thank you for the tips , with your and MaxHoek's advice I found a better workflow today for working from my low poly to my hight poly, and realized the importance of control loop for it.
Wirrexx : since its a a game asset I'm trying to optimize and get used to model with the least geometry !
@Nizidranamitz thanks, and nice to hear that you could improve your skills. But @wirrexx and @perna are right. If the pieces are in rl different objects, do also model them like this. Probably for lp you could model them in one piece. But for hp you always have to model things in different objects, like they are in rl. Thats very very important, keep it in mind!
And the problem with the pinching is because of your bit messy geo i think. Try to keep polys in one size. Also think about you model before you start in 3d. "Is this one piece or not?" " How much geo do i need in my cylinder?" ETC.
This is a simplified example of the general problem of sub-d modeling intersections between complex smooth surfaces and hard edged surfaces. - I know that having more original segmentation before subdividing helps preserving the original curvature of he smooth part of the resulting surface. - I know that we can clenaup a boolean or use a hihg-res boolean to do a retopo job on it. - I know that we can do a high-density dynamesh in zbrush and let it retopologize automatically.
But if we put those techniques away and stick to badass polymodeling, what would be the best solution? I found out that actually leaving a 5-gon when doing the support edges produces least of the well-known artifacts. Any thoughts on using n-gons like this?
Thanks. P.S. I used to be a fanatic quad-maniac. I'm not sure if that's right.
@aandronov i cant see the problem there. its very basic for hp modeling. If you want, you dont need any manually placed supporting edges. just setup you smoothing groups and use the double smooth method. The problem here is also, put in more geo and dont work with this shitty supporting edges on such a round surface.
@aandronov i cant see the problem there. its very basic for hp modeling. If you want, you dont need any manually placed supporting edges. just setup you smoothing groups and use the double smooth method. The problem here is also, put in more geo and dont work with this shitty supporting edges on such a round surface.
Sure thing, but this always means to propagate way too much subdivisions over broad smooth areas that simply don't need it.
@aandronov i cant see the problem there. its very basic for hp modeling. If you want, you dont need any manually placed supporting edges. just setup you smoothing groups and use the double smooth method. The problem here is also, put in more geo and dont work with this shitty supporting edges on such a round surface.
Sure thing, but this always means to propagate way too much subdivisions over broad smooth areas that simply don't need it.
Not really. You need this geometry for such shapes. So you have the choice. Work with more geo from the start, or work with more geo with your turbosmooth. Depends on what you want to achieve afterwards.
This is a simplified example of the general problem of sub-d modeling intersections between complex smooth surfaces and hard edged surfaces. - I know that having more original segmentation before subdividing helps preserving the original curvature of he smooth part of the resulting surface. - I know that we can clenaup a boolean or use a hihg-res boolean to do a retopo job on it. - I know that we can do a high-density dynamesh in zbrush and let it retopologize automatically.
But if we put those techniques away and stick to badass polymodeling, what would be the best solution? I found out that actually leaving a 5-gon when doing the support edges produces least of the well-known artifacts. Any thoughts on using n-gons like this?
Thanks. P.S. I used to be a fanatic quad-maniac. I'm not sure if that's right.
being lazy. but here's my "try" left top -> Box left top 2:nd -> 2x2x2 Box left top 3rd -> Turbosmooth x2 + spherify left top forth -> inset and extrude Bottom left -> Turbosmooth with smoothing groups Bottom Right -> Turbosmooth x2 iterations
Perna Was talking about using geo that already exist! Because my boxsphere had enough segments for me to use, i could create the shape coming out easy and using the existing geo as support!
IMO people really need to get on the OpenSubdiv train. It's a far superior way to model hardsurface assets that allows you to still utilize support edges if necessary, while reducing complexity overall and being much quicker to work with. It also gets you in the habit of modeling clean and uniform topology that is easier on the rendering side of things, and the only downside is that you have to do final edge weighting at the end of the modeling process. But wait! That means that instead of mucking around with unneeded and cumbersome support loops, you can tweak edge sharpness however you like! So if an art director came back and told you to loosen or tighten or change some edges, it's so trivial you could make the change and have the AD clear the asset on the spot.
Yeah, OpenSubdiv and crease sets are good, but the results are not optimized due to the same reason - geometry is propagated along all surfaces. Traditional edgeloops still give you the option to distribute mesh resolution. Plus, creasing is quite similar to double-smoothing with smoothing-group-separation in the first Turbosmooth iteration. Without the important ability to use different weights, of course. Anyways, my question concerns old-school-support-edge modeling.
Yeah, that's what I initially meant by having more original edgeloops/segmentation. However, if I want a really tight edge to pop out of a really broad smooth surface, I'd have to propagate more and more geo in all directions.
somedoggy: OpenSubdiv in the implementations I've seen is essentially just the ancient double subd approach. As such you have next to zero control over the end result. The second you model out something more complex than what you show there, you're in trouble. Maybe something has changed since I last had a look at it, but it was pretty awful back then. Wouldn't be able to use it successfully for say a gun or a similarly complex mesh, not to mention that there are all sorts of issues with export/import and even maintaining the edge data internally.
With quadchamfer we're beyond support loops already, and can still control our edge widths to perfection.
I use it almost exclusively now and have not experienced any trouble making complex forms. Give me a small hardsurface challenge if you'd like! As for import/export I haven't had to work much with interoperability to be fair, but for game assets it works no different with baking at least since you export it as a pure, collapsed mesh. I assume it works fine going from Max to Maya, so if you can stick to them I see no problems using this in production. I'm not sure what other use cases there are in games for a non-collapsed highpoly. Maybe you'll have more ideas in that regard. Also, if you're bringing it into Zbrush it would give you a cleaner, more uniform model to work with than regular subdiv so that's a nice plus.
@aandronov This is exactly where I mentioned that you still use loops if they are the right option. This is the kicker with Opensubdiv. You relegate loops to where they work best and you can be a lot more liberal with them, keeping geo more consistent. You have also made a mistake in your usage. Notice that your edges in the OpenSubdiv version are not actually smoothed out at all compared to the original boundary mesh. This means that your subdivision level was too low for the edge weight value you used. Modeling with OpenSubdiv requires different thinking and will not work if you just try to slap it on without understanding first. If I want to sharpen these edges more I can either: 1) Increase the edge weighting at the cost of needing to subdivide more. 2) Push the support edges out more at the cost of less polygonal density uniformity.
Both are valid and together create an incredibly flexible technique.
You have also made a mistake in your usage. Notice that your edges in the OpenSubdiv version are not actually smoothed out at all compared to the original boundary mesh. This means that your subdivision level was too low for the edge weight value you used. Modeling with OpenSubdiv requires different thinking and will not work if you just try to slap it on without understanding first.
I made a polycount comparison, that's why I stopped at a relatively lower subdiv level on the OpenSubDiv. I'm perfectly aware that on higher subdiv level it will look OK. I agree equal density is a better starter for Zbrushing. I agree that combining different methods is best in almost all cases. (e.g. OpenSubDiv doesn't totally replace support-edges technique, which I tried to illustrate)
Would anyone be interested in seeing a shape attempted using OpenSubdiv rather than more traditional techniques? Perhaps one of the weekly hardsurface challenges? I figure it'd be a useful example of a different approach, where little examples currently exist on polycount afaik.
Would anyone be interested in seeing a shape attempted using OpenSubdiv rather than more traditional techniques? Perhaps one of the weekly hardsurface challenges? I figure it'd be a useful example of a different approach, where little examples currently exist on polycount afaik.
Sure thing, I'd be most interested! It would be best If you could demonstrate both sharp and semisharp corner usage in one model.
@perna its like every workflow i think. it depands on the shape you want to achieve and also the personal way to model things. Im stuck with the standard support edge modeling in combination with double smooth with smoothing groups. I think "standard" is the most flexibel and cleanest method. but sometimes its easier with double smooth to get more geo on some shapes. but like almost, it depends on personal preferences i think. but i dont have any experience in the industry, so i`m happy to be here to learn and hear what others have to say. How do you model most of your things @perna ? and maybe more important, why do you model with this technics?
Max: I've never seen double subd used in a professional environment, and I've seen a lot of artists. I've only seen it used as a crutch, on the polycount forums, due to lack of modeling knowledge. That may not be the case with you - it would be interesting to see where you feel it's a more worthwhile approach than loops (automated or not).
Oh okay. Because I learned this method in university from some hardsurface artists at crytek. I also use a script from one guy there, so you can easiely change smoothing groups to speed up this process. I know there are many many loops who are worth it, but the pcs right now are stong enough to handle this. I love this method because you can model very low res with only some good placed support loops to get a good result. The crytek guys told me so much tricks there.
I mostly use QuadChamfer. For some projects I've made assets without a single manually placed control loop. You practically get your hipoly for free. I separate meshes by material. Plastic, metal and rubber will be separate.
I never heard from this method. lol! do you have a good guide or advice for this? would be very nice!
QuadChamfer is a maxscript, that max users had to download to compensate for the sloppy job the traditional chamfer tool in 3ds max did. Luckily, Autodesk included similar functionality in max2015. It basically does support edges for free in many cases.
@perna Can you elaborate on not being able to control edge widths? I do know the importance of being able to define edge sharpness but have not experienced any problems doing this via OpenSubdiv.
I'm all for using whatever method to get the best result, but I'm not sure I can agree on all uses of double smooth being just as a crutch... If we're talking about the same thing. Turbosmooth with smoothing groups on + another Turbosmooth? Off the top of my head, Ben Mathis uses it and I've seen it used on The Order. Sometimes it's a quick option that looks fine IMO, as long as you always do it with control.
As for OpenSubdiv I believe that if Pixar artists can create gorgeous hardsurface work like WALL-E, it's not a matter of the tool being inferior. It's a matter of not using it well, like a crutch just as you've said.
This is a quick example I mocked up. Not trying all that hard to specifically stick to the original so don't judge that side of things too harshly, but are these not controlled edge sharpnesses? I specified them exactly how I wanted, and if I thought an area could be improved it'd be trivial to weight it differently. No "melting" on my end, as far as I can tell. There is one problem area I can pick out but I also know how to clean it up. Again this is more of a demonstration than anything.
I'd like to hear more of your thoughts!
Edit: There's two main problem areas with this mesh actually, both easily tweakable to get the exact edge profile you'd want.
Max: I've never seen double subd used in a professional environment, and I've seen a lot of artists. I've only seen it used as a crutch, on the polycount forums, due to lack of modeling knowledge. That may not be the case with you - it would be interesting to see where you feel it's a more worthwhile approach than loops (automated or not).
I've used it (and creasing) and have seen multiple co-workers use it. I don't think of it as a crutch if you're getting work done much more quickly as a result.
Obviously, having precise edges is ideal; however, I feel that quick deadlines and quick revisions don't allow for it all the time.
@perna That's totally fair to go for such control, especially when aiming for machined edges. However what's stopping you from combining this with OpenSubdiv? They don't appear mutually exclusive to me, and I already work with both depending on how I want to produce my edges. As long as you know where to put the topology it should produce the same quality. I've done work like that for some companies (I think an iPhone model I made is damn near accurate down to the millimeter).
I don't use the QuadChamfer modifier, but sometimes I do want to vary my edge sharpnesses with good control, not based on material but have it varying across the object. If I want them as accurate as your examples I just chamfer the edges myself. I see this as very useful over something more global as parts can be milled any which way and won't always be a single chamfer width. Is this something the modifier supports as well?
Now I don't think though that people are necessarily trying to get accuracy to life 100% of the time, especially for more organic hardsurface like carved wood for example. And certainly not when going for stylization. Being pragmatic I still can't say it's a crutch to use double smooth in production, namely when you have to create something novel and actually know what you're doing. Can I do it pure and perfect? Sure, no problem. In the end though the most readable thing is the sharpness of the curve and if you can control that look well enough with another technique like double smooth or OpenSubdiv then hey cool. With OpenSubdiv though, unless there's a hole in my logic, I can still use chamfers to create perfect edges where I like OR take a more softcore approach and weight edges, all on the same mesh. This is not something that a simple double smooth can do with anywhere near the finesse, like the soft weighting on the front of my model versus the sharp weighting on the top capsule piece.
P.S. Was it you who posted a guide on perfect edge chamfering for subd a long time ago? I vaguely remember a post about how to make sure edges don't lose their rounding but I can't find it for the life of me on Google. It showed a few various edges with support loops on a 90 degree angle that result in a more sloppy edge than if you were to chamfer it and use support loops. Something like this:
Below is your design but made with QuadChamfer. It's at worst the same amount of work but with a better result, more accurate, faster, more optimized, more easily tweaked, so why choose the inferior approach on purpose? You'll note that the edges are perfectly uniform, a result you won't achieve with double subd. Lookdev is so easy since the mesh updates in realtime as the chamfer value spinner moves, and once I have the look I want I can instance that QuadChamfer modifier to all other objects in the scene which use the same material. That sort of consistency and efficiency was just a game changer when this modifier was released (a long time ago now).
Could you please send me your .max file perna? i never heard about this method and i would like to learn how it works. I also searched for your video with pedro, like you mentioned, but dont found it hehe
Could you please send me your .max file perna? i never heard about this method and i would like to learn how it works. I also searched for your video with pedro, like you mentioned, but dont found it hehe
Same here (about the video, looked through YT but couldn't find anything). I'll be able to switch back to Max in a month or so would definitely like to explore the workflow beforehand. I will say though that the method is very dependent on 3ds Max and the modifier stack. In Maya there are way fewer options and I would be curious which method is the quickest in that case.
Huh, I don't want to derail this thread too much but yea this looks really solid. Thanks for bringing it up again.
Edit: I should probably head to bed now but I ended up hoping back into Max and gave it a quick try. The areas in blue had some issues though. Pretty neat though. I'll have to dive in deeper when I have some more time.
thanks man, good stuff. I also watched it a few weeks ago. If learned very much from this video, thx! But there is no real desciption how you are using quad chamfer at all. Did you always select every edge before adding the chamfer modifier, or did you do this with smoothing groups? i tried it with sg and no manual tweaking at all, but there are big artifacts. Do you know what im doing wrong? do i also need to tweak after chamfer by hand for this model?
first time im doing this, so dont judge me please :DD
The quad chamfer is not doing very well at the red marked spots. would be nice if someone could give me a quick advice for this. the method sounds really good in your words. I will definitely give it a try!
thanks man, good stuff. But there is no real desciption how you are using quad chamfer at all. Did you always select every edge before adding the chamfer modifier, or did you do this with smoothing groups? i tried it with sg and no manual tweaking at all, but there are big artifacts. Do you know what im doing wrong? do i also need to tweak after chamfer by hand for this model?
first time im doing this, so dont judge me please :DD
I selected the edges. The reason for that was.
1. I have a smaller Hole and 2 bigger. Giving them the same amount of bevel/quad, will give the smalle one a sharper edge when zooed out. . 2. I had more control over the places that needed fixes afterwards. Just where you've marked you'r stuff.
For example the box that sticks up, you've got a 6 vertices on one face, mine is only a quad. Saves a lot of time doing it that way than 2 add support loops yourself
It worked pretty good i think. but @perna you told that you could change your edge width also afterwards. In my case it doesnt work. Maybe i need to prepare my mesh perfectly, before adding the champfer to change it afterwards. But i dont know, i think it doesnt work in some cases. Or am i wrong? because i know how good your models look and how long your doing that,so i think you know exactly what you are talking about xD
I also did this model with my current modelling method:
here you have also the chance to change the edge width afterwards. Not so flexible you told before, but also enough if you think about your model before.
To change edge width, you just change the chamfer value?
But if i change it after i cleaned the mess of the chamfer it will fuck up my mesh. So i asked you: 1. clean up after the chamfer or... 2. no clean up, just put the chamfer on the perfect prepared mesh
You're adding support edges to a double subd mesh... doesn't that defy the entire point of...
And yes, you need this. otherwise, the mesh will not look very clean. you only need a few. to define the edge width. I dont know if you understand the method im talking about correctly. thats what i learned from very good artists at crytek ffm. dont think they do a bad job there.
They told me this method with the backgorund: work with less polys than possible, place good supporting edges and let the programm (turbosmooth in this case) do his work.
Somedoggy, at this point you're being defensive of a technique purely on the basis that it is the technique you use, not its technical merits. You ignore my points and claim that double subd can do things that it demonstrably can not (You have zero edge width control with double subd/OpenSubdiv, stop trying to claim otherwise). I don't see that as constructive. I'm trying to help. Take it or leave it, but please don't waste my time.
The only people here who defend double subdiv are people who don't use QuadChamfer. That says a lot.
Yes, I've made a lot of posts on chamfering. You may be referring to the YouTube video I made with Pedro.
I was having a discussion. Please don't perceive my talking points about a different method as defensiveness. The only thing I hard-line disagreed with is that double smooth =/= OpenSubdiv. Your example was good and I appreciate you sharing your methods. No time wasted. And thank you for pointing me back to that video.
Using support edges on double smoothing actually is useful in certain circumstances. Obviously the main goal of double smooth is to just say screw it and let max do all the work but if your blockout is more dense in a certain area and you happen to have a large quad of nothing in one area, you can compensate that area with some support edges (But DO NOT place them as near to your intended edges as it will over-tighten that area, rather just spread it out if possible). You can sort of use the double smooth method with support loops too to "control" the tightness of the edges to a certain degree but I have never really done that before, but IT IS possible.
Somedoggy, by defensiveness I mean how you hold on to ideas after they've been proven false and don't respond to my counter-arguments. As for your hard-line disagreement, "=/=" means NOT equal, so you are attributing to me a stance opposite to the one I made. I assume you just meant "=".
Guys, don't just make claims. Demonstrate them: A - Show a mesh which is significantly faster to make with double subd than with QuadChamfer. B - Show how the result of OpenSubdiv differs from that of double subd C - Show how you can achieve consistent edge widths with double subd without violating [A]. D - In fact show why anyone except a person who is not very good at subd modeling would ever double subd.
... I mean as in actual images of actual meshes. You can't make game art with jibber-jabber. I've demonstrated my points with images, so it should be easy for any other 3D artists to do the same. Otherwise it's only talk, innit?
BARKIT BREK
Good points, I've given this a good bit of consideration and I'll see if I can't get the modifier to do tests. And yes I meant to say "=".
Max, I believe your understanding of this stuff is fundamentally broken and needs to be re-learned more or less from scratch. Don't dismiss this suggestion on the basis of its harshness, you're really breaking major rules and common sense associated with the use of these techniques.
No no @perna its fine to get critics from somebody. Im happy to be here, so somebody can kick my ass and not just talking its fine what im doing. But would be better if there is some constructive criticism and not just you need to start from scatch xD You also need to know. I never heard about modeling with the chamfer modifier, so its for me like to learning to ride a bike and just fallen down . I love to learn more and im very ambitious to get better, but you also need to understand that no one is born a master. Its the first time i used that technic and i know its just a trick to get supporting edges more or less "automatically". Nothing more and nothing less than that. Its just a cool tipp, but before you need to understand the overall subD modeling. I completly agree.
But like @supaclueless very well written, there are some supporting loops, because its no dynamesh after putting on the first turbosmooth. For sure just spare placed, but they need to hold the shape. And i would say there are shapes you can model quicker with double subd, but im done with this discussion and also not good enough to talk in detail about such thing. There are other people out here who could probably do this.
Indeed, the 3rd party script offers extended functionality, but I recently moved to max2016 and currently I don't have QuadChamfer installed. I only wanted to see the resulting wires of the middle/right shape (before meshsmoothing) or a max file with collapsed mesh.
I just wanna chime in here and say - I've been double smoothing for the past 2 years or so ever since i first tried it, sped up my workflow like crazy. But after reading the discussion going on in this thread right now i decided to finally drop the cash on quadchamfer to try it out, and wow. @perna you weren't exaggerating. Suffice to say that 2nd turbosmooth is never, ever coming back
The wires are pure traditional subd control edges. No difference whatsoever. No fancy approach, no trickery, nothing clever.
Buy a license, it's terribly cheap. (Please don't warez that thing, support small devs) Then invest time in mastering the different options. In my view being spoonfed with exact scenes and parameters and so on it detrimental to learning. Much, MUCH better to go in clueless and learn every aspect of something through trial and error. That way the lessons stick, and you will be on another level of understanding, able to do a lot more.
Agreed, I always do so, but I had the false impression that you wanted to demonstrate some specific trickery. Since that's not the case wires are pointless. Thanks.
Ah yea, with quad chamfer there is the option to connect out those things. Granted I'm still using Maya at work so I'm just experimenting at night. So far though it's definitely a neat toolset that I'm looking forward to exploring more and seeing how it fits into my pipeline.
Now here's a technical question for you guys to cheer you up a bit! It really bothered me many times today. Does anyone happen to know if there's an option in 3ds max that makes the program crash in the most inappropriate moments and is enabled by default? If yes, please let me know how to turn it off. Thanks.
Replies
Got these square extrude of a cylinder, did a chamfer. Would like to turbosmooth but I have some really weird thing happening (screen). If I add control loop near those edges, its end up (logically) having a hightlight, which I dont want.
How do I manage that please? Do I have to separe these pieces of the cylinder?
Here is what i mean:
Here is what i came out with:
I`m working with two turbosmooth and smoothing groups, for less support edges. Maybe you could also try it for shapes like this: http://polycount.com/discussion/117488/double-smooth/p1
Thanks for the tips ! I was already using double smooth though, but lacked the supportive edge at right spot, now its better.
Because of the extrude my edge flow wasn't really good and I had to do a new mesh with highter geo before extruding those part. Then I did double smooth and it worked way better, although I Still have some pinching... maybe i'm still missing some geo?
@perna
Hello Perna, I'm sorry you guys have to answers again and again the same questions, I didn't knew that my cases could be issues with Wirrexx example.
Thank you for the tips , with your and MaxHoek's advice I found a better workflow today for working from my low poly to my hight poly, and realized the importance of control loop for it.
Wirrexx : since its a a game asset I'm trying to optimize and get used to model with the least geometry !
And the problem with the pinching is because of your bit messy geo i think. Try to keep polys in one size.
Also think about you model before you start in 3d. "Is this one piece or not?" " How much geo do i need in my cylinder?" ETC.
This is a simplified example of the general problem of sub-d modeling intersections between complex smooth surfaces and hard edged surfaces.
- I know that having more original segmentation before subdividing helps preserving the original curvature of he smooth part of the resulting surface.
- I know that we can clenaup a boolean or use a hihg-res boolean to do a retopo job on it.
- I know that we can do a high-density dynamesh in zbrush and let it retopologize automatically.
But if we put those techniques away and stick to badass polymodeling, what would be the best solution?
I found out that actually leaving a 5-gon when doing the support edges produces least of the well-known artifacts.
Any thoughts on using n-gons like this?
Thanks.
P.S. I used to be a fanatic quad-maniac. I'm not sure if that's right.
Not really. You need this geometry for such shapes. So you have the choice. Work with more geo from the start, or work with more geo with your turbosmooth. Depends on what you want to achieve afterwards.
left top -> Box
left top 2:nd -> 2x2x2 Box
left top 3rd -> Turbosmooth x2 + spherify
left top forth -> inset and extrude
Bottom left -> Turbosmooth with smoothing groups
Bottom Right -> Turbosmooth x2 iterations
https://www.dropbox.com/s/i3f93vezfq82d5b/test.max?dl=0
Test 2
"cleaner mesh"
https://www.dropbox.com/s/gux9vdqrf0m1alh/test2.max?dl=0
Perna Was talking about using geo that already exist! Because my boxsphere had enough segments for me to use, i could create the shape coming out easy and using the existing geo as support!
Yeah, OpenSubdiv and crease sets are good, but the results are not optimized due to the same reason - geometry is propagated along all surfaces. Traditional edgeloops still give you the option to distribute mesh resolution. Plus, creasing is quite similar to double-smoothing with smoothing-group-separation in the first Turbosmooth iteration. Without the important ability to use different weights, of course.
Anyways, my question concerns old-school-support-edge modeling.
@wirrexx
Yeah, that's what I initially meant by having more original edgeloops/segmentation. However, if I want a really tight edge to pop out of a really broad smooth surface, I'd have to propagate more and more geo in all directions.
Thanks.
OK, I don't deny any merits of the OpenSubDiv approach, but to continue the argument look at this:
Just look at the edge distribution:
Greetz,
Andy
Sure, I had to recreate it as I did not save the file the first time:
https://drive.google.com/open?id=0B4vo_o3utgw6WU83OEp5TDFWWjA
@aandronov
This is exactly where I mentioned that you still use loops if they are the right option. This is the kicker with Opensubdiv. You relegate loops to where they work best and you can be a lot more liberal with them, keeping geo more consistent. You have also made a mistake in your usage. Notice that your edges in the OpenSubdiv version are not actually smoothed out at all compared to the original boundary mesh. This means that your subdivision level was too low for the edge weight value you used. Modeling with OpenSubdiv requires different thinking and will not work if you just try to slap it on without understanding first.
If I want to sharpen these edges more I can either:
1) Increase the edge weighting at the cost of needing to subdivide more.
2) Push the support edges out more at the cost of less polygonal density uniformity.
Both are valid and together create an incredibly flexible technique.
I agree equal density is a better starter for Zbrushing.
I agree that combining different methods is best in almost all cases. (e.g. OpenSubDiv doesn't totally replace support-edges technique, which I tried to illustrate)
@perna
Thanks, that answered my question. 5-gon usage is not uncommon and produces acceptable results in this particular situation.
Would anyone be interested in seeing a shape attempted using OpenSubdiv rather than more traditional techniques? Perhaps one of the weekly hardsurface challenges? I figure it'd be a useful example of a different approach, where little examples currently exist on polycount afaik.
It would be best If you could demonstrate both sharp and semisharp corner usage in one model.
Thanks.
its like every workflow i think. it depands on the shape you want to achieve and also the personal way to model things. Im stuck with the standard support edge modeling in combination with double smooth with smoothing groups. I think "standard" is the most flexibel and cleanest method. but sometimes its easier with double smooth to get more geo on some shapes. but like almost, it depends on personal preferences i think.
but i dont have any experience in the industry, so i`m happy to be here to learn and hear what others have to say.
How do you model most of your things @perna ? and maybe more important, why do you model with this technics?
I never heard from this method. lol! do you have a good guide or advice for this? would be very nice!
Luckily, Autodesk included similar functionality in max2015.
It basically does support edges for free in many cases.
I'm all for using whatever method to get the best result, but I'm not sure I can agree on all uses of double smooth being just as a crutch... If we're talking about the same thing. Turbosmooth with smoothing groups on + another Turbosmooth? Off the top of my head, Ben Mathis uses it and I've seen it used on The Order. Sometimes it's a quick option that looks fine IMO, as long as you always do it with control.
As for OpenSubdiv I believe that if Pixar artists can create gorgeous hardsurface work like WALL-E, it's not a matter of the tool being inferior. It's a matter of not using it well, like a crutch just as you've said.
This is a quick example I mocked up. Not trying all that hard to specifically stick to the original so don't judge that side of things too harshly, but are these not controlled edge sharpnesses? I specified them exactly how I wanted, and if I thought an area could be improved it'd be trivial to weight it differently. No "melting" on my end, as far as I can tell. There is one problem area I can pick out but I also know how to clean it up. Again this is more of a demonstration than anything.
I'd like to hear more of your thoughts!
Edit: There's two main problem areas with this mesh actually, both easily tweakable to get the exact edge profile you'd want.
Obviously, having precise edges is ideal; however, I feel that quick deadlines and quick revisions don't allow for it all the time.
I don't use the QuadChamfer modifier, but sometimes I do want to vary my edge sharpnesses with good control, not based on material but have it varying across the object. If I want them as accurate as your examples I just chamfer the edges myself. I see this as very useful over something more global as parts can be milled any which way and won't always be a single chamfer width. Is this something the modifier supports as well?
Now I don't think though that people are necessarily trying to get accuracy to life 100% of the time, especially for more organic hardsurface like carved wood for example. And certainly not when going for stylization. Being pragmatic I still can't say it's a crutch to use double smooth in production, namely when you have to create something novel and actually know what you're doing. Can I do it pure and perfect? Sure, no problem. In the end though the most readable thing is the sharpness of the curve and if you can control that look well enough with another technique like double smooth or OpenSubdiv then hey cool. With OpenSubdiv though, unless there's a hole in my logic, I can still use chamfers to create perfect edges where I like OR take a more softcore approach and weight edges, all on the same mesh. This is not something that a simple double smooth can do with anywhere near the finesse, like the soft weighting on the front of my model versus the sharp weighting on the top capsule piece.
P.S. Was it you who posted a guide on perfect edge chamfering for subd a long time ago? I vaguely remember a post about how to make sure edges don't lose their rounding but I can't find it for the life of me on Google. It showed a few various edges with support loops on a 90 degree angle that result in a more sloppy edge than if you were to chamfer it and use support loops. Something like this:
Could you please send me your .max file perna? i never heard about this method and i would like to learn how it works. I also searched for your video with pedro, like you mentioned, but dont found it hehe
Huh, I don't want to derail this thread too much but yea this looks really solid. Thanks for bringing it up again.
Edit: I should probably head to bed now but I ended up hoping back into Max and gave it a quick try. The areas in blue had some issues though. Pretty neat though. I'll have to dive in deeper when I have some more time.
SCENE IS UNDER.
https://www.dropbox.com/s/m3pj03n4i3lt0f4/myresult.max?dl=0
i tried it with sg and no manual tweaking at all, but there are big artifacts. Do you know what im doing wrong? do i also need to tweak after chamfer by hand for this model?
first time im doing this, so dont judge me please :DD
The quad chamfer is not doing very well at the red marked spots. would be nice if someone could give me a quick advice for this. the method sounds really good in your words. I will definitely give it a try!
1. I have a smaller Hole and 2 bigger. Giving them the same amount of bevel/quad, will give the smalle one a sharper edge when zooed out. .
2. I had more control over the places that needed fixes afterwards. Just where you've marked you'r stuff.
For example the box that sticks up, you've got a 6 vertices on one face, mine is only a quad. Saves a lot of time doing it that way than 2 add support loops yourself
It worked pretty good i think. but @perna you told that you could change your edge width also afterwards. In my case it doesnt work. Maybe i need to prepare my mesh perfectly, before adding the champfer to change it afterwards. But i dont know, i think it doesnt work in some cases. Or am i wrong? because i know how good your models look and how long your doing that,so i think you know exactly what you are talking about xD
I also did this model with my current modelling method:
here you have also the chance to change the edge width afterwards. Not so flexible you told before, but also enough if you think about your model before.
If you want, here is also my max file:
https://www.dropbox.com/s/rzhwnrcb4nwh0ra/myresult.max?dl=0
1. clean up after the chamfer
or...
2. no clean up, just put the chamfer on the perfect prepared mesh
But than:
And yes, you need this. otherwise, the mesh will not look very clean. you only need a few. to define the edge width. I dont know if you understand the method im talking about correctly. thats what i learned from very good artists at crytek ffm. dont think they do a bad job there.
They told me this method with the backgorund:
work with less polys than possible, place good supporting edges and let the programm (turbosmooth in this case) do his work.
You also need to know. I never heard about modeling with the chamfer modifier, so its for me like to learning to ride a bike and just fallen down . I love to learn more and im very ambitious to get better, but you also need to understand that no one is born a master. Its the first time i used that technic and i know its just a trick to get supporting edges more or less "automatically". Nothing more and nothing less than that. Its just a cool tipp, but before you need to understand the overall subD modeling. I completly agree.
But like @supaclueless very well written, there are some supporting loops, because its no dynamesh after putting on the first turbosmooth. For sure just spare placed, but they need to hold the shape.
And i would say there are shapes you can model quicker with double subd, but im done with this discussion and also not good enough to talk in detail about such thing. There are other people out here who could probably do this.
Could you please share an image with wireframe of your QuadChamfer setup or the scene file for the example above?
Thanks.
Indeed, the 3rd party script offers extended functionality, but I recently moved to max2016 and currently I don't have QuadChamfer installed.
I only wanted to see the resulting wires of the middle/right shape (before meshsmoothing) or a max file with collapsed mesh.
Thanks.
Gonna shell it out for QC.
Does anyone happen to know if there's an option in 3ds max that makes the program crash in the most inappropriate moments and is enabled by default? If yes, please let me know how to turn it off. Thanks.