Polyhertz's 3dsMax script - 'Quick High-Poly'
musashidan
high dynamic range
From an ongoing discussion on this thread - http://polycount.com/discussion/164254/maya-surface-based-asset-generation#latest - in which @Fansub first demoed his Shapeshifter tool for Maya - which us Max users drooled over - it turned out that @Polyhertz mentioned, humbly and nonchalantly 
, that he had created a script for Max that ran a similar operation..........and he did indeed.
So last night he sent me the script and I got to test it out today. And here are the results below. The script basically adds an inset based on smoothing groups and automatically opens the quad chamfer dialogue. I think the results speak for themselves. They are results that simply cannot be achieved with quad chamfer alone. This script to quad chamfer is as spinach to Popeye!
@PolyHertz , I think it would be well worth your while polishing this, adding a few features(inset preview top of the list) and flogging it on Gumroad. I'm sure it would be a huge success.
This was an example I recreated from the Fansub thread that noted the fact that quad chamfer just could not deal with this topology.
As stated by Fansub himself, these tools are not magical wands and some cleanup for optimal results will always be required, especially with boolean topo.
Note the shading artifacts from the boolean operations. Cleanup was a piece of piss in prep for the script.
Pre-chamfer inset based on smoothing groups is the secret sauce.
This was another Fansub mesh design that I copied to demo the script in comparison to Shapeshifter directly.
And finally, I took it through the workflow in under an hour(including making 2 cups of tea and going out for a smoke or 2) - Boolean mesh unwrapped and triangulated before export. There are a few seams that needed addressing but I didn't bother for the demo.
Sorry @Fansub I didn't do your mesh justice but I just paused your youtube vid and eyeballed it rather quickly.
, that he had created a script for Max that ran a similar operation..........and he did indeed.So last night he sent me the script and I got to test it out today. And here are the results below. The script basically adds an inset based on smoothing groups and automatically opens the quad chamfer dialogue. I think the results speak for themselves. They are results that simply cannot be achieved with quad chamfer alone. This script to quad chamfer is as spinach to Popeye!
@PolyHertz , I think it would be well worth your while polishing this, adding a few features(inset preview top of the list) and flogging it on Gumroad. I'm sure it would be a huge success.
This was an example I recreated from the Fansub thread that noted the fact that quad chamfer just could not deal with this topology.
As stated by Fansub himself, these tools are not magical wands and some cleanup for optimal results will always be required, especially with boolean topo.
Note the shading artifacts from the boolean operations. Cleanup was a piece of piss in prep for the script.
Pre-chamfer inset based on smoothing groups is the secret sauce.
This was another Fansub mesh design that I copied to demo the script in comparison to Shapeshifter directly.
And finally, I took it through the workflow in under an hour(including making 2 cups of tea and going out for a smoke or 2)
- Boolean mesh unwrapped and triangulated before export. There are a few seams that needed addressing but I didn't bother for the demo. Sorry @Fansub I didn't do your mesh justice but I just paused your youtube vid and eyeballed it rather quickly.
Replies
-
Looks great!
I would definitely check that out.
-
Gad! That's incredible. So those sharp edges and ngons are not a problem for baking in substance painter? Looks like I will be hiking over to gumroad to throw money at this script when it becomes available.
Good job on the demo info btw.
Cheers
I luv the tech forum. -
Interesting thanks, for sharing your demo musashidan.
-
Now i'm popular hahaha
@PolyHertz you did an amazing job on this one ! And musashidan the explanations you've put up are simply marvelous ! Do you mind if i use your explanations as a template to document ShapeShifter ?
Really in love with what i'm seeing here !
EDIT : For those of you curious to see this workflow in action,I've made a quick video to showcase the whole process,from a cube to a baked mesh in Marmoset Toolbag 2.
https://www.youtube.com/watch?v=Q7h-rDei4A4
-
@Fansub, Ah that clears it up. Great work!
-
@fansub And I'm sure you'll get more popular.Fansub said:Now i'm popular hahaha@PolyHertz you did an amazing job on this one ! And musashidan the explanations you've put up are simply marvelous ! Do you mind if i use your explanations as a template to document ShapeShifter ?
Really in love with what i'm seeing here ! Yes, of course you can use them. I'd be honoured. And now I'm off to watch your new video again. Haha!
-
Yes, definitely worth checking out.floon said:Looks great!
I would definitely check that out.
-
@agitori You're welcome, mate.
-
this has made my freakin day. Just a question though. How does it deal with subdivision? Afaik quad chamfer doesn nice chamfers on complex geo as well but when It comes to applying turbosmooth, it fails / needs cleanup.
-
I need this.
-
The idea of this workflow is that you're creating very complex-looking models that look turbosmoothed without the need to do any time-consuming sub-d modeling. These methods leave a lot of poles, even on clean topo, so it isn't designed for sub-d for that and the reason above. The examples above are all extremely quick/easy to create this way, but imagine the time it would take to sub-d model them.(not to mention much more complex models like AAA FPS guns, for example)Bumblebee said:this has made my freakin day. Just a question though. How does it deal with subdivision? Afaik quad chamfer doesn nice chamfers on complex geo as well but when It comes to applying turbosmooth, it fails / needs cleanup. -
By any chance you might know when this is going to be released? Much thanks!!!
-
It's great script.
-
looks very very good! definitly want to check it out! good job
-
When is it going to be released!? I need this hardly. Thanks!
-
This looks absolutely fantastic, thank you for posting it @musashidan! Do you know where Polyhertz's Max script is located? I can't seem to figure out where it is. Maybe I just need more coffee.
Edit: Definitely need more coffee. I misread the topic! @PolyHertz - I'd gladly pay to support the development of this tool! -
@perna The script acts as a barrier that 'pushes' back the many n-gons created by the boolean geo. It adds an inset based on smoothing groups before the chamfer operation. So it essentially reroutes the shitty topology into a clean inset. Why this hasn't already been added as a feature of the chamfer mod I'll never.......oh yeah, Adesk....
@Synaesthesia Cheers mate. Just read your edit.
For all those wondering about the script. I didn't write it, Polyhertz did. I'm just testing it out for him. He says he wants to see how Fansub's Shapeshifter tool for Maya goes(see post link at page top) before he does anything with it. Hopefully he'll respond here when he sees the thread.
-
This looks great! Even if sub-d doesn't apply well to it, looks like it's a perfectly fine mesh to take into Zbrush and run dynamesh on if needed.
-
Quad chamfer alone + turbosmooth does stuff like thisperna said:I feel pretty dumb right now. Someone explain why you can't just use quad chamfer for these models?
Ignore the horrible chamfer and bad example just threw this together
Not really sure as to why you inset, really the only problems are caused by Ngons not being connected well, which is solved using turn to poly
If it's to avoid the Ngon issue like shown, just make sure you have an edge to support the geo somewhere, also quadchamfer has an inset option within edit chamfer option.
I'm not trying to make the script any less valuable btw, but I can see why some are confused. -
My reply to perna above explains it. Try it yourself on a complex booleaned mesh, QC alone just can't handle it.Mossbros said:
Not really sure as to why you inset, really the only problems are caused by Ngons not being connected well, which is solved using turn to poly
-
@musashidan
I get it, but I'm not seeing said results when trying the method of insetting by smoothing groups, then chamfering the smoothing groups, so something else must also be happening within the script I assume?
As when you break down into those steps and try to do it manually, you get bad smoothing no matter what.
Well, at least I do with anything moderately complex.
-
Hey guys! Wow at this response.
I didn't realize people wanted this so much, I mean it's not that complex of script really. For those that are asking for copies of it, the one I sent @musashidan still needs some work before being ready for release. I'll finish it up and post a copy on Gumroad soon.
@Mossbros Turbosmooth/MeshSmooth absolutely hate complex N-Gons like in your example. Like you've observed, inseting and chamfering don't fix this basic issue. It -might- be possible to do something about this on completely flat n-gons, but I'm not quite sure yet.
I don't want to give anyone the wrong impression; the script will not make turbosmooth work with complex n-gons any better then it already can. This may change later on, but atm that's just not something it's built to fix.
-
@Mossbros yes I'm sure there's more going on behind the scenes in the script but the inset before chamfer is the trick it seems. The script is encrypted so I really don't know Polyhertz's secret.
-
That's the whole ideainstg8r said:This looks great! Even if sub-d doesn't apply well to it, looks like it's a perfectly fine mesh to take into Zbrush and run dynamesh on if needed. a complex sub-d looking mesh without the sub-d ing. -
Ah, the man himselfPolyHertz said:Hey guys! Wow at this response.
I didn't realize people wanted this so much, I mean it's not that complex of script really. For those that are asking for copies of it, the one I sent @musashidan still needs some work before being ready for release. I'll finish it up and post a copy on Gumroad soon. I had a feeling this would be popular based on the response Fansub got for Shapeshifter. Now get it polished! -
perna said:I appreciate your position, @perna , but I don't see the harm in having multiple options for creating high-poly models. If the final result is a success and it gets done on time then what does it matter the approach used?
I've done the 'hard yards', I've been modeling sub-D for over a decade, learning the difference between an N-pole and an E-pole back when sculpting didn't exist and Subdivision.com was the place to be.
I can understand your concerns for the lack understanding the fundamentals, the lazy approach, the amount of people I see now learning 3D with the 'how can I do this as fast as possible with as little effort as I can muster' or 'I don't like unwrapping or retopo, can't I just press some buttons that will do it for me?'
This method certainly isn't ideal for everything, but suits certain assets. The same way sub-D modeling will remain the flagship method for years to come. It's definitely not going anywhere in VFX work.
When I say complex I mean it in the sense that anybody who knows their way around a set of sub-d tools knows. The traditional topological intersections that have given many a man nightmares. You know what I mean. The 'ugly mess' part doesn't matter. This is an aesthetic mesh.....just make sure wireframe is turned off.
If the bake/render is clean then it's a result.
And the method shown here still requires some understanding of topology to troubleshoot.
-
perna Dunno if you've already seen it but i made a timelapse video using this method on a relatively complex mesh.The video originally took 21 minutes but i sped it up to 7 minutes.
I agree with you on these points : not only is this method over-hyped (the amount of clients i had that expected SS to make their high polies is crazy) but you also can't beat that good ol' powerful subD in terms of flexibility.
With that said,some props like the one in my video are imho easier and faster to do with this workflow or the Zbrush one.When the only thing you expect from subD is to smooth your corners i do think other solutions can sometimes be useful,but anything that requires complex shapes is a no-go atm.
The goal here is just to smooth the corners of a mesh like the Zbrush method,anything past that i always do it in subD.
-
This is clever. If the inset by smoothing group step could be implemented as a Max creation graph, rather than a script, it could be added to the stack for a non destructive workflow. In that way, you could build your stack up all the way including the boolean, for a totally interactive preview.
Stack might look something like this:
Quad chamfer
[inset by smoothing group]
Smooth
ProBoolean
Final meshes would likely still require topo and smoothing group fixes and tweaks, but the interactivity could be beneficial. -
Wouldn't doing a double chamfer get the same results as inset by smoothing group? (Well not the same, but similar.)
Chamfer - Tension: 0.5 Segments: 4 From Smoothing: Unsmoothed
Chamfer - Tension: 0.0 Segments: 1 From Smoothing: Unsmoothed
Edit Poly
I tried it on Fansub's demo model and it looks very similar to the screenshot above.
-
while this is awesome, on a second thought I do agree. People call me obsolete but to me turbosmooth is still a essential modifier. Simply because I like to have the "optimal" geometry. So for instance, on some pieces I use very low geometry because 1. its fast 2. easier to work with 3. changes made quickly and this is because turbosmooth does the refinings. Using this approach you re kinda forced to pump a decent amount of geometry in some cases and like perna or OP said, cleanup is still a thing. There are alot of people out there who kind of forget what subdivision actually is for. They pump so much geometry in pieces that could be done with at least 70% less geo. Long story short ; although im impatiently waiting for the release, I am not a huge fan of workflows you have to pump tons of geometry into for it to look good.perna said:Musashidan, you seem to be saying some things which are unrelated to my post and attributing claims to me which I have never made. I make two explicit claims, both of which can be addressed unambiguously:
- There is no significant advantage to using the proposed method (smoothing group inset before chamfer) over manually placed edges.
- The method is guaranteed to become a crutch for artists who believe it will do work for them that they themselves can't be arsed to learn to do using traditional means.
Fansub, I had a look at the video. A few manually added edges should be sufficient to prep that hipoly cage mesh for QuadChamfer. I can demonstrate that if you upload the mesh. As you agree, the method is over-hyped. That's absolutely not criticism of your efforts or toolset. You know what it is capable of doing. It seems that a lot of people here don't, however.
A lot of artists here struggle with producing clean boolean results. A tool which validates such results will likely cause these artists to spend much more time learning to do things properly than they otherwise would have. That would have no negative effect on me whatsoever. I'm trying to help people avoid shooting themselves in the foot here.
-
I am not a huge fan of workflows you have to pump tons of geometry into for it to look good.Keep in mind that this workflow is for high polies,and no matter what gymnastic you do a mesh using this technique will almost always have a lower polycount than the same mesh with subD,assuming that the subdivision was applied on the mesh.
Imho what lakes on this workflow is just proper tools and experimentations,but hey this is just a start.If people keep playing with it ideas will come to life and we'll be able to improve this method in the future,which is great imho.
I insisted a lot on the non-production-ready of this method when i first made it,and that is still valid as of this day.But if you can pump a prop with it and have it nicely baked don't hesitate to use it for a second as long as you know your shit.
-
Yes, aware of that. But I refer to working with optimal geometry. Its true the topology is irrelevant on highpolies but I am not talking about the actual polycount but the functionality. Also, my statements are in a context of why one would prefer this over quad chamfer.Fansub said:Keep in mind that this workflow is for high polies,and no matter what gymnastic you do a mesh using this technique will almost always have a lower polycount than the same mesh with subD,assuming that the subdivision was applied on the mesh.
Imho what lakes on this workflow is just proper tools and experimentations,but hey this is just a start.If people keep playing with it ideas will come to life and we'll be able to improve this method in the future,which is great imho.
I insisted a lot on the non-production-ready of this method when i first made it,and that is still valid as of this day.But if you can pump a prop with it and have it nicely baked don't hesitate to use it for a second as long as you know your shit.
-
Also, my statements are in a context of why one would prefer this over quad chamfer.Maya user here so no idea about how QC operates in this situation
I am not talking about the actual polycount but the functionality.What's the "functionality" you're referring to,out of curiosity ? If i'm working on my personal sci fi scene and have the ability to pull multiples props with this technique,why would i care about the functionality as long as i know what i'm doing and pumping the expected result ?
-
I kinda see where's perna coming from - this won't help much when it comes to modeling real-life stuff where all the fillets are slightly different, there are some sharp edges and some soft ones etc. For example this:Fansub said:What's the "functionality" you're referring to,out of curiosity ? If i'm working on my personal sci fi scene and have the ability to pull multiples props with this technique,why would i care about the functionality as long as i know what i'm doing and pumping the expected result ?
Though about 90% of the time spent here was not spent modeling it but rather comparing the reference images and the blockout to get it right, which leaves a grand total of 3-5% of time spent on mesh cleanup.
For concepts, sci-fi scenes and kitbashing, your tool makes sense, although quick auto-remesh + subdivision (or relax on top) might be just as fast and in some cases even more foolproof. -
How do I purchase that script for 3ds max?
-
I kinda see where's perna coming from - this won't help much when it comes to modeling real-life stuff where all the fillets are slightly different, there are some sharp edges and some soft ones etc.Then i think we all agree on this this inset+bevel technique was made for specific cases where the user would only need one uniform bevel,the second you go into the grounds of versatile bevels that nicely increase/decrease their width in certain areas SubD just comes back and knocks the door.
For people who don't know how to model properly these tools aren't gonna help just like the rounded edge or proboolean workflow didn't help,but for others these techniques can be very useful at creating smooth angles really quickly.
-
This tool might lock you in to a specific bevel, but you can do all of this manually and get the same result. The question is whether or not you'd like to have more manual control over your output. I can see it being a really useful way to work to get high-poly models created in a very short timeframe, as opposed to traditional subdivision which can take considerably longer as the model's complexity increases. Skill in each method will certainly affect this, but I think overall using this technique will certainly improve speed if that's what you're after.
-
I'm sorry you feel that way @perna I thought I adequately addressed your points and certainly had no intention of putting words in your mouth. In fact, I actually agreed with you somewhat.perna said:Musashidan, you seem to be saying some things which are unrelated to my post and attributing claims to me which I have never made. I make two explicit claims, both of which can be addressed unambiguously:- There is no significant advantage to using the proposed method (smoothing group inset before chamfer) over manually placed edges.
- The method is guaranteed to become a crutch for artists who believe it will do work for them that they themselves can't be arsed to learn to do using traditional means.
I think some people may be 'over-hyping' indeed. Every time something like this pops up there is an initial frenzy. I'm not advocating 'the death of modeling fundamentals' or encouraging laziness. I'm merely testing a tool and sharing the results. And the reason I posted those results here is because I had hoped for a discussion......and here we are.
My personal philopsophy has always been to learn as much as possible; as many tools, methods, and workflows as I can; and throw out what stagnates me whilst keeping the useful stuff. Every time I aproach a model, no matter how simplistic or complex, my brain goes through the options I have readily available. Will I use sub-d, will it be with in-line edges or will I also use creasing/opensubdiv, straight polymodeling, splines or primitives, will it be box-modeled or edge-extruded, can I sculpt and retopo, booleans, dynamesh sculpting, procedural modeling, or the many ways I have developed in Zbrush that work for me. Only a fool would look at a method like the one discussed and think it's the holy grail of lazy modeling. It isn't. It's just another option.
And yes, I wholeheartedly agree that an understanding of clean boolean/solid modeling is extremely lacking. But that's what I stated above, that a knowledge of topology is still a requirement to achieve efficient results with this method, whereas the lazy who can't be arsed in the first place to learn the fundamentals will inevitably fail. Each artist will decide. Those who have no options to turn to either learn or perish.
-
Yes, @monster it is very similar, but the results from your double chamfer are not as clean. I ran it on the same default demo mesh with your settings. In this example it's clear to see the 'barrier' effect that the inset has on the n-gons.monster said:Wouldn't doing a double chamfer get the same results as inset by smoothing group? (Well not the same, but similar.)
Chamfer - Tension: 0.5 Segments: 4 From Smoothing: Unsmoothed
Chamfer - Tension: 0.0 Segments: 1 From Smoothing: Unsmoothed
Edit Poly
I tried it on Fansub's demo model and it looks very similar to the screenshot above.
-
Well, you can adjust the numbers a little to match pretty close. Disabling "Limit Effect" avoids the waves that form, and is safe to do when the Chamfer Amount is small. But in the end it looks like the inset leaves slightly cleaner result because it's adding another row of quads. I don't have the script so I can't really test but I'm sure it'll come out soon.
-
I assume you're all talking about Max's new built-in Quadchamfer? I can't get a double-stacked QC modifier to pass on smoothing groups in Max 2014 (using Mario's modifier)..
-
where can i get the script ?
-
@cptSwing Yeah, in Max 2015+. You just uncheck Smooth Result at the bottom for it to retain smoothing groups. Also, it's the same old Chamfer modifier with Quad as an option. I was confused and was looking for "Quad Chamfer" at first.
@David_T It's not released yet. musashidan is testing the script for PolyHertz.
-
This looks very good, but , I have a question , that perhaps is not directly related to the script but to the models i am seeing, they look cleanand smooth but from the tutorials I been following they all say to avoid Ngons , hows that those are used there even if flat survace? I thought that those superpoly faces shoudl be always avoided in modeling and just use quads everywhere?
-
from the tutorials I been following they all say to avoid Ngons...Imho people say this either because they want you to first learn how to model without Ngons or they're still stuck in the stereotypes of the past.
Ngons are not bad at all if you know how to use them,and because this thread is about experimentation on workflows with Ngons,having complex shapes with weird ngons is only going to help us figure out the strengths and weaknesses of this method
-
@perna Just to follow on from what I was saying about this just being another tool in the box that is for every artist to decide when to use it or not. Here is a model that I had to create yesterday, and I had decided on using the method here for one of the parts completely independantly of what we've been dicussing. Yes, I could have used sub-D for the part but I only had half the day to model/UV/bake/texture it, so I made a decision as I thought that this method would suit the part perfectly. It was quick to do, the bake turned out OK and I got a result I was happy with. So I suppose it's a good example of when it is a useful method. But let me reiterate, I do agree with you on this becoming a crutch for beginners who couldn't be arsed putting in the sub-d man hours and learning about topolgy, but in reality they are only setting themselves up for failure. Unless they have a solid understanding of topology, tri-edge orientation effects on surface shading, smoothing groups, efficient boolean operand placement, etc, then this method will be frustrating and ineffectual.
Of course the asset is nothing special or complicated but I think that the forms of the blade are suitable for this method when time is limited, there is no blockout in place, and you know it's going to work before you start.
High poly
-
This is pure polygon modeling. The 'dreaded' n-gon, and what you might have heard pertains to sub-divisional modeling. Quad based sub-d has long been drummed into peoples' heads. The reason is that they perform more predictably(mathematically) when ran through the smoothing algorithm. Tris and n-gons still subdivide to quads(3x4=12/5x4=20/6x4=24/7x4=28/etc) but the >=6 valence poles are where the trouble occurs. This is where 6 or more edges converge at a single vert. A good example of this is to subdivide a cylinder with a polar cap. This is most noticeable on curved surfaces but can be imperceivable on flat surfaces due to the way the surface is shaded.NAIMA said:This looks very good, but , I have a question , that perhaps is not directly related to the script but to the models i am seeing, they look cleanand smooth but from the tutorials I been following they all say to avoid Ngons , hows that those are used there even if flat survace? I thought that those superpoly faces shoudl be always avoided in modeling and just use quads everywhere?
This argument has raged for as long as sub-d has been around(a long time in CG terms) and the bottom line is: understand the theory behind topology and you don't even need to worry about it. You will just know what to do.
Quads are still the most efficient solution to use as they lend well to modeling(clean edgeloops make it easier to control the topology), UVing is easier, rigging is easier. All based on the fact that quads create continuous loops and are easier to deal with generally.
And remember, in the context of game asset creation(which this method is suited for in some cases) the topology of the final in-game mesh will be triangulated so knowing how to deal with it is important. As @perna said quite correctly, you can't just ham-fistedly brute force the topo and expect clean results. Prep work, cleanup, and a clear understanding of what you're doing is vital.
