Ngons do not matter one bit when using the Boolean to Dynamesh workflow, that's a the whole point of it.
As for hard ops : it is interesting, but extremely limited since it does not allow for any tweaking/shape shifting after the fact, which is what the technique discussed here is all about.
Ha, this I don't know, as I haven't compared the output of both. That said I haven't even bothered looking at the meshing of this example model - I just applied a triangulate modifier and exported, and it behaved as expected when Dynameshed
Ngons do not matter one bit when using the Boolean to Dynamesh workflow, that's a the whole point of it.
Zbrush can have issues with n-gons as they are 'illegal topology', which can sometimes cause issues with dynamesh. I usually just enable the 'make quadrilaterals' option in proboolean.
Yes, that's precisely why I mentioned I applied a triangulation modifier before exporting
At the end of the day, for this technique to be useful the user shouldn't have to worry about meshing at all.
I thought that once the primitives were down in Blender that the segment count couldn't be edited? That would limit the ability to get a fast LP out, unless I miss how to do that (very possible, my Blender experience is limited).
'Tis a shame Autodesk don't just grab the open-source code for "Instant Meshes" and finally integrate a Dynamesh solution of their own. Getting to be a bit of an embarassment.
This seems like a great way to produce some complex models. The only thing I feel that I would struggle with is producing the low poly model. I guess using a tool like polydraw in Maya would do the trick?
Eh, I dunno — without live primitives you're severely limited when it comes to revisions. Sure, you can save backup meshes at various stages but honestly, it's a clumsy workaround that only gets more restricted as complexity increases. And when it comes time to do the lowpoly and you realise your original backup piece for the lowpoly counterpart needs to change, you're out of luck.
Having said that, one advantage of working destructively is you learn to commit to changes rather than umming and ahhing and wasting time — this is ideal for personal work when you're the one making the calls, but when you can't predict what changes may need to be made at any point down the road, it's hard to say no to automating tasks and simplifying major changes.
Thanks for sharing the technique, an eye-opener to me! I have forgotten ProBoolean since I was learning modelling ~5 years ago, but this is a major boost to hard-surface workflow. Much better than dynameshing objects directly in Zbrush, as its non-destructive; better than modo`s Mesh Fusion, as it`s simpler, doesn`t crash so much and you can sculpt details on top of dynameshed tool.
Sounds good on paper, but if you work in a destructive workflow in Modo
you got some decent scene management tools to duplicate a mesh before
making important changes, got a max limit of 500 undo actions or more,
save incrementally often, etc. In 3DS Max, you won't use modifiers all
the time and if you do you'll end up collapsing it at some point. The
modifier stack is nice, but it's not as relevant as some people like to
claim. You'll find several Modo users on Polycount who were once are
still are 3DS Max users and still find the destructive modeling workflow
in Modo more efficient (implying that the lack of a modifier stack is
not an issue). But if a user has been modeling exclusively in 3DS Max
for years and has never really experienced a destructive workflow
similar to Modo, then I can understand why some people think that the
modifier stack is objectively better when others think it's partially
superficial. And for those who still are firmly convinced that it is an
issue, then be happy, because Modo will soon have similar features for
non-destructive workflows.
I've been modeling in Max for donkey's years now, and have also done a modeling project or 2 in Modo. For me personally I rarely work with an edit poly mod in the stack as the keyboard mismatch is a pain in the arse. Stacking and collapsing often is usually my workflow.(symmetry/shell/turbosmooth/FFD being my most common combo) Yes, I can certainly live without the stack for straight modeling, but for other things like this thread's topic, the stack is extremely valuable.
And as for a destructive workflow being more efficient? I can't see the logic in this. Backing up files, cloning meshes 'for a rainy day', and setting your undo buffer to 500 doesn't exactly equate to efficient in my eyes.
Blender booleans also don't automatically quadify some of the resulting faces so after a few steps of booleans you'll get drastically higher vertex counts than you would in Max with PB. It's still a really good way to get a highpoly in Blender but you'll be stuck doing retopology or a lot of cleanup afterwards. If you need a quick lowpoly with Blender and your hard surface doesn't deform (which is, I mean, pretty common for a hard surface) I guess you could just set Decimation Master to one percent on your dynameshed model and just use that for your lowpoly, maybe with a tiny bit of cleanup in some areas. It might look a little weird to you but I bet 99% of players aren't going to be able to tell the difference.
Great write-up! I've heard of the workflow before, but this really simplifies it for me. Going to have to give it a try on my next project. Thanks for taking the time.
For me personally I rarely work with an edit poly mod in the stack as the keyboard mismatch is a pain in the arse.
That's a glaring problem in Max that I'm surprised they never fixed. It's worth the time to go through both Editable Poly and Edit Poly, unify your hotkeys so you don't have to think about it.
I've written scripts that switch out commands based on the selection context, and I run all my hotkeys through those now. It's a pain to set up but worth it.
Just FYI, as @musashidan mentioned. Max 2017 Boolean has been updated and should solve faster than ProBoolean now. It also has a streamlined UI and a Boolean Explorer.
I have heard of this workflow before but never actually used it (didnt understand it properly). But now I´m convinced that this can be a great timesaver on certain hard-surface props. I just tested it and this is really cool.
I've written scripts that switch out commands based on the selection
context, and I run all my hotkeys through those now. It's a pain to set
up but worth it.
I suppose I'd just gotten so used to avoiding edit poly mod and using editable poly that I didn't give much thought to a workaround, buy definitely good to know. thanks for the info/link.
Just remembered that I used this exact method - proboolean>dynamesh - on this project I did this time last year having watched the Joe Drust vids. The structure that the character resides in is built this way.
Just FYI, as @musashidan mentioned. Max 2017 Boolean has been updated and should solve faster than ProBoolean now. It also has a streamlined UI and a Boolean Explorer.
Big downfall though is that with the old-school Boolean you can't extract the operands, plus it lacks the advanced options of Proboolean. Pity they didn't overhaul Proboolean instead.
Also, I encountered a lot of buggy behaviour, although in fairness I was testing it in the beta.
Well, you shouldn't need Extract Operand anymore since Boolean now just hides the selected objects. But it is missing a few features. I just like the new Boolean Explorer workflow.
I'll stop talking about this because I don't want to take away from the awesome technique being discussed here.
How do you avoid visible edges on your HP after the process? ProBoolean with really high edge counts?? I used like 64 segments on some of these cylinders.
Yep, he mentioned this in 1st post:
One thing to note - keep your curves dense at this stage. When using cylinders, use high segment counts. I often use 36 at minimum and 108 or 140 for extremely large cylinders. This will produce better results in Zbrush. Later, when generating the lowpoly, we'll dial down these segment counts. That's easy to do because they remain live subobjects within the Proboolean.
Chris Rohlfing: Oh man, I don't even want to think about making a low poly model of my engine block. But yeah, polydraw would be ok. I think I'd rather just try to Zremesh it, and unwrap based on a planar angle threshold. Quick and dirty. I don't think it's necessarily worth doing something like that manually.
Thanks for sharing, its a very interesting workflow that i cant wait to try out. I was wondering on people options on rather using Proboolean, using Multi mesher scrips i came across not too long ago.
I haven't tried the boolean method just yet but i wonder if the multi meshers might give a better visual feedback on the progress of the overall shape and save you needing to undo if the shape is not just right. It also seems to add topology to your mesh while your working. Its true that it may not give you the best topology result but its still something.
Nice technique ! I use the modals pie menu on blender to make my assets with mirrors, booleans, arrays and in the end I use my clean face to delete ngons.
did another test, this time I took it to zbrush, dynamesh high setting, slight polish pass, did a zremesher pass of around 300K polys. Then brought back in Max , mesh was cleaner and did need to optimize. Still only a High poly.
That looks great @juniez. Max has a few 3rd-party remeshing plugins but I don't know a way to smooth them out in a similar way to Zbrush's polish. It looks like Blender can handle all of that internally.
Both have its merits. Knowing proper subd modeling allows you to be more versatile if you plan to move away from games and into film or cinematics where subd is important because certain renderers tessellate meshes based on distance to the camera.
I wouldn't discard any method for building assets. Always good to have an extra method of doing things in case stuff needs to be made faster.
Now that the method described above seems to be praised by most of us, does that mean that we finally started to care more about productivity than the quality of high poly models?
Every professional I know wants to balance quality with
efficiency. That's nothing new. It's why threads like this get front page, why people buy
tutorials, and why we have a tech talk forum.
Even though unlike max you can not alter primitives after creating them anymore the lowpoly should not be a problem to achieve because you never have to collapse the stack on the operands. So approaching the boolean with halfway subd capable models that double up resolution-wise as good bases for game output and then using subdiv modifiers on all of them for the desired roundness of the high res output is doable.
Then you have to run the boolean op without the modifiers and clean up the result.
During past years, we used to see comments from expert subd modelers that in order to get professional high quality subd models, each hard edge had to look nice and unique, something we can't get from a generic smoothing method such as the one used in the workflow described above. Personally, I think productivity is far more important on a game production than the quality of a high poly model that will lose quality/details anyway through the baking process once projected as a normal map on a game asset.
Now that the method described above seems to be praised by most of us, does that mean that we finally started to care more about productivity than the quality of high poly models?
Don't forget that using this technique, you're ending up in Zbrush, with a sculptable mesh, and an incredible set of sculpting brushes. You can sculpt. Edges can be customised if needs be using any number of tools. You can set up polygroups and polish by groups, you can Trim dynamic angled bevels, use the clipping brushes, mask out different corners and run a stronger/weaker polish for softer/harder edges.....
And yes, the way I see it, this technique is perfect for a game-bake-purpose mesh as we don't give a s@@t about the topo or UVing, it's simply a case of aesthetics. Also, illustration renders and portfolio images will obviously benefit.
But in VFX/film/etc, as @Pedro Amorim mentioned, meshes will often be straight-up polygonal geometry with production-ready topology. Some of these assets could have 50 UDIM tiles so UV unwrapping is a big consideration. A lot of VFX artists don't bake, but paint displacement maps.
Yes, this would be fine for background assets/etc but hero models will always have production-quality topology.
So I wouldn't set aside your knowledge of the difference between an E-Pole and an N-Pole just yet. Sub-modeling/high-res polygonal modeling still has its place.
Another important factor, I feel, is that with the rising popularity of VR, polygonal modeling could become more important than baking in a lot of cases. Hero assets could completely shift to single-asset geometry rather than normal map baking for silhouette purposes. This would obviously render the high-poly obsolete. Apparently NM detail doesn't translate well to VR due to the nature of the viewing experience. Plus the ever-rising vert count limits factor.
Of course, this is just a personal theory. None of us know what'll happen. More strings to the bow, I say. Learn all the viable techniques you can. "Be like water, my friend"
Here is a test using Blender's modifiers (mesh must be triangulated before): It works pretty good, but there will be some performance problems with more complex meshes.
@McGavish Yes, this is essentially the same as quadify mesh>turbosmooth in Max. As @Amsterdam Hilton Hotel mentioned earlier, not so great with more complicated meshes.
Import boolean constructed object as voxels (or make an object with 3d Coat primitives and its boolean ops on geometry) and then Geometry->Smooth All I would assume.
Import boolean constructed object as voxels (or make an object with 3d Coat primitives and its boolean ops on geometry) and then Geometry->Smooth All I would assume.
Yes exactly, just import and scroll down the menu on the left side till commands->"smooth all" press it a few times and then export.
Replies
As for hard ops : it is interesting, but extremely limited since it does not allow for any tweaking/shape shifting after the fact, which is what the technique discussed here is all about.
how do you texture that last crazy mesh though. seems very uv-unfriendly. or do you just use it for normals for another mesh?
Nice tutorial. Thanks for sharing.
At the end of the day, for this technique to be useful the user shouldn't have to worry about meshing at all.
Definitely. the whole idea is automation, proceduralism, non-destruction, and......relief from bloody tedious topology noodling!
but i would go the "old" way if i have to match referece images or fotos...
Indeed indeed - certainly an issue if one wants to generate a quick low out of that.
This seems like a great way to produce some complex models. The only thing I feel that I would struggle with is producing the low poly model. I guess using a tool like polydraw in Maya would do the trick?
Having said that, one advantage of working destructively is you learn to commit to changes rather than umming and ahhing and wasting time — this is ideal for personal work when you're the one making the calls, but when you can't predict what changes may need to be made at any point down the road, it's hard to say no to automating tasks and simplifying major changes.
That's a glaring problem in Max that I'm surprised they never fixed. It's worth the time to go through both Editable Poly and Edit Poly, unify your hotkeys so you don't have to think about it.
I've written scripts that switch out commands based on the selection context, and I run all my hotkeys through those now. It's a pain to set up but worth it.
Per went deep into this topic years ago in possibly his best thread
I have heard of this workflow before but never actually used it (didnt understand it properly). But now I´m convinced that this can be a great timesaver on certain hard-surface props. I just tested it and this is really cool.
Thank you!
http://polycount.com/discussion/168276/styx-zbrush-maya-max-tool
ProBoolean with really high edge counts?? I used like 64 segments on some of these cylinders.
Just remembered that I used this exact method - proboolean>dynamesh - on this project I did this time last year having watched the Joe Drust vids.
Also, I encountered a lot of buggy behaviour, although in fairness I was testing it in the beta.
Use more edges or keep repeating the polish until they go away.
I'll stop talking about this because I don't want to take away from the awesome technique being discussed here.
http://www.scriptspot.com/3ds-max/plugins/multi-mesher
I haven't tried the boolean method just yet but i wonder if the multi meshers might give a better visual feedback on the progress of the overall shape and save you needing to undo if the shape is not just right.
It also seems to add topology to your mesh while your working. Its true that it may not give you the best topology result but its still something.
Anyone's thoughts?
I use the modals pie menu on blender to make my assets with mirrors, booleans, arrays and in the end I use my clean face to delete ngons.
Cheerio
I wouldn't discard any method for building assets. Always good to have an extra method of doing things in case stuff needs to be made faster.
This is a tool to add to our belt. It is not completely replacing the belt. If it can make production faster, awesome.
But this'll never be quite as impressive as a nice clean wire, oof.
Even though unlike max you can not alter primitives after creating them anymore the lowpoly should not be a problem to achieve because you never have to collapse the stack on the operands.
So approaching the boolean with halfway subd capable models that double up resolution-wise as good bases for game output and then using subdiv modifiers on all of them for the desired roundness of the high res output is doable.
Then you have to run the boolean op without the modifiers and clean up the result.
And yes, the way I see it, this technique is perfect for a game-bake-purpose mesh as we don't give a s@@t about the topo or UVing, it's simply a case of aesthetics. Also, illustration renders and portfolio images will obviously benefit.
But in VFX/film/etc, as @Pedro Amorim mentioned, meshes will often be straight-up polygonal geometry with production-ready topology. Some of these assets could have 50 UDIM tiles so UV unwrapping is a big consideration. A lot of VFX artists don't bake, but paint displacement maps.
Yes, this would be fine for background assets/etc but hero models will always have production-quality topology.
So I wouldn't set aside your knowledge of the difference between an E-Pole and an N-Pole just yet.
Another important factor, I feel, is that with the rising popularity of VR, polygonal modeling could become more important than baking in a lot of cases. Hero assets could completely shift to single-asset geometry rather than normal map baking for silhouette purposes. This would obviously render the high-poly obsolete. Apparently NM detail doesn't translate well to VR due to the nature of the viewing experience. Plus the ever-rising vert count limits factor.
Of course, this is just a personal theory. None of us know what'll happen. More strings to the bow, I say. Learn all the viable techniques you can. "Be like water, my friend"
It works pretty good, but there will be some performance problems with more complex meshes.
http://lesterbanks.com/2016/03/klaudio-ladavacs-editable-boolean-tools-for-maya/
Yes exactly, just import and scroll down the menu on the left side till commands->"smooth all" press it a few times and then export.