[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vr / w / wg] [i / ic] [r9k] [s4s] [vip] [cm / hm / lgbt / y] [3 / aco / adv / an / asp / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / qst / sci / soc / sp / tg / toy / trv / tv / vp / wsg / wsr / x] [Settings] [Home]
Settings Home
/3/ - 3DCG

Thread archived.
You cannot reply anymore.


File: checker_large.gif (3 KB, 256x256)
3 KB
sup /3/

I want to talk about procedural and parametric workflows. I work in Archviz and so far I've done everything in Rhino and Grasshopper. Rhino on it's own is pretty shit desu but Grasshopper is amazing. But now I'm working with UE4 a lot and Rhino is pretty useless for that. The mapping tools are shit, the mesh tools are shit... you can kinda make it work with GH but it does not feel very efficient.

So I'm looking for more CG oriented software, but it should also have these procedural and parametric possibilities of Grasshopper. So far I can only think of 3dsmax + railclone or Houdini. I know neither of both so idk which one I should learn. Or maybe I'm missing something else?

Anyone here integrated these in your workflow and can shed some light on issues or just tell your experience? Maybe someone else is using Rhino and UE4 and can shed some light on the issues of mapping, smoothing groups etc...

If you work in Archviz, did you try some other plugins like VisualArq?

What problems do you run in using Rhino to UE4?
substance designer is ontop of the food chain so to speak.
but i used to make some procedural in cycles like wood & metal. simply not anymore because it lacks grunges and it only have patterns
File: rhino wat r u doing.jpg (226 KB, 1869x955)
226 KB
226 KB JPG
tried VisualArq for a bit, but it wasnt very reliable imo.

the problems are mainly related to uv mapping and mesh control. Rhino is a nurbs modeler made for designing things to produce irl. So when you try to use it for CG you run into issues. The meshing algorithms are not very great, leading to very fucked up topology. Mesh editing tools are pretty much non existent. Things like bevel or chamfering are very inconvenient to do..

The mapping tools are very basic as well. I managed to do a decent enough script for mapping surfaces but unwrapping anything that's not a box is a pain in the ass. Then there are no smoothing groups. Instead, Rhino has vertex welding. For example, if you have two mesh faces that are connected, they will each have their own vertices. If you want an edge to be smooth you have to weld the vertices, but UE4 does not recognize this as smoothing groups. It renders it regardles, but it's a pretty shit way to deal with this. Especially since there are no ways to select edge loops or something like that.

Today, I mapped a pillar I made and then used "match mapping" on the 919 other pillars. It did not work, for some reason only some random faces were mapped on the other pillars. When I imported the pillars into Maya, it was like viewing the UV mapping from the top... really weird. I mean, UV maps are normally 2D, right? It was like they were rotated towards the screen. Like laying a sheet of paper on the table and then looking at it from the side. Wtf.

No matter what you do, you always have to go through Maya or Max or another CG suite if you want smoothing groups or hard eges but without having 2-3 times the amount of vertices..

Oh and by the way, rhino and GH both can't into multi threading. It just feels very outdated for what we're trying to do and I'm growing pretty frustrated with it.

pic related
created a surface mapping with GH and in the UV editor, the shells are somewhere in the boonies. It worked fine in UE4 but wtf man
substance is amazing but I'm looking for software to actually create and modify geometry

and there's such an abundance of materials out there that it has not been necessary for us to use substance for now
you mean procedural geometry? im not sure what you mean.
you can displace geometry with a modifier just like im showing here but you can also displace geometry for render only purpose and have variety of options in cycles.
still, the best options would be to get a displacement/height map from substance designer, but if you want to do it quickly you can punch in some random map in maya/3ds and it can do that too
File: rheno.jpg (247 KB, 1710x980)
247 KB
247 KB JPG
no, I mean like this:

>get 2D drawing from architecture department
>30k m2 shopping mall
>need to make about 1000 pillars
>select all of the 2D squares supposed to be pillars
>reference them in my Grasshopper script
>set a height and a chamfer angle
>hit bake
>I now have 1k pillars with chamfered edges, turned into a nice clean mesh with smooth edges and a proper mapping
>import into UE4
>colleague comes downstairs, tells me client made some changes and pillars are now thicker and positioned differently
>repeat first steps, have new set of pillars in about 2 mins

Basically, creating geometry that remains procedural, or maybe parametric is a better word. This works in rhino but it has a lot of issues so I'm looking for something where I can also do what I described, but with a more up to date package and more tailored towards CG instead of irl production.
oh, that's a completely different game. you mean engineering sketches.
procedural sketches? not yet. this is far too complex. maybe try to input this in revit and see what you get. i don't come from architecture so i wouldn't know
>That UI node
what planet is this from ?
Revit and the likes are upstream in the pipeline, and unfortunately my office has not jumped on the BIM train yet.

What node do you mean?
I think Houdini would be the ticket here, they have a plugin for UE4 that makes integration rather simple: https://www.sidefx.com/products/houdini-engine/ue4-plug-in/
I just googled and it seems people actually use Rhino to convert .dwg files into a Houdini format.

So maybe I can get the best of both worlds, the versatility of Grasshopper for semi-automatic generation of geometry and then Houdini for fleshing it out.

Anyone know how Houdini handles UV mapping of procedural objects?
you can uv procedural fine. houdini is all procedural so anything you see made in it its good lol. its an incredibly intense software, you can make practically anything mathematically possible, if you learn VEX/VOPS (google it). and mantra ( the renderer) is very good, it also comes with renderman.

that being said rhino and grasshopper are still used in the industry. i have very little experience with them so i cant give advice but whatever problem you're having are certainly workable.
I'm not going to drop Rhino+Grashopper completely, because there are so many other situations where they are useful. Just frustrated with the pipeline to UE4. It's certainly possible but sometimes it feels like eating soup with a fork.

I played around with Houdini in the last hour and already having fun, gonna dive into it tomorrow. Gonna be a steep learning curve but I hope that I will eventually be able to create UE4 assets like railings or girders and what not.
Would be amazing to have control over their size and shape inside UE4. It happens quite a lot that changes are made during production and making them right in the engine would be very convenient.

Could you recommend any tutorials for a beginner looking to create architectural UE4 assets?
there were a bunch of gdc+siggraph talks about using houdini for game assets that were pretty essential to me understanding the workflow. its especially nice because they assume you have a prior understanding of other software and are just showing how houdini is helpful in the pipeline and what its capable of




basically id structure your learning based on getting a good idea of how things are built in a general sense then learn the specific nodes after. itll speed you up significantly, esp when looking through the manual
File: IMG_4055.jpg (23 KB, 325x266)
23 KB
I eat cardboard boxes can someone made me a 3D model of Thomas the tank engine with a bonsai tree growing out of his head. Make it lewwwwwwwd. Thanccccccxxxxxx is /3/
substance is like a little baby compared to Houdini, yes, even in texturing.
Yes this is childs play for Houdini, with HDA's you can even make these changes right in UE4.

I doubt you need me telling you this, but make sure you give apprentice a good workout and test this pipeline out so you know what kind of issues you will run into ahead of time, you will run into issues, but they will be solvable!
Thank you so much, much appreciated.
I'm hoping that some of the GH knowledge will carry over. I'm sort of familiar with node based workflows and in GH it was really easy once I understood the structure.

Is there anything like a forum where people can discuss or get advice?
For GH there is the grasshopper3d forum where you can get tons of plug ins and pretty much any problem you could run into has been solved before, and most people simply share their definitions for free. I saw there was some sort of Houdini asset store, is that comparable?

Thanks, already downloaded it. Really awesome that they have this apprentice version and not the usual student or trial stuff. Sidefx seems like a very cool and laid back company.

As for the pillar problem I described, things like that happen all the time. My Archviz company is associated with an architecture firm and rather than visualising finished designs, we come in right at the start and work with the architects on the design. We're sort of their real time rendering plugin. So we have to adapt to constant changes and GH has been doing a good job, but now with UE4 we could really use something more CG oriented.

Almost sounds too good to be true to be able to make changes right in the engine, or with little overhead in Houdini. I hope I can make this work.
theres the sidefx forums and i think some threads on polycount and a fairly inactive subreddit. none are super active but you can usually get an answer

if you really want to get your feet wet Anastasia Opara has a massive 5 volume gumroad tutorial creating this lakehouse. one of the best houdini resources ive found (although pretty expensive, the first 3 are on cgp tho)

its easy to get overwhelmed tho, so if you're like me and tend to skip the really simple beginner stuff try not to. modelling a rope or something may seem boring but its essential to getting it to click

just watched the first video here >>572173
and it's pretty much the same I would do in GH to create such a bridge. Only easier. Amazing.
Since I literally only know Rhino+GH, I would start with beginner stuff anyway. I did see that Anastasia stuff beforehand and I really wanna get there at some point. Doesn't really matter how expensive that is if it helps the company in the long run.
good luck on your journey
That's grasshopper, a plugin for Rhino
looks terrible and the construction makes no sense

It's free btw

You do have a point. But you can play around with some sliders and make 50 different versions in 2 minutes...
Thanks for the reply.

Yeah, UVs are an absolute nightmare with Rhino, but I've seen the rise of textures that don't need UV mapping anymore. Triplanar mapping I think they were called. Wouldn't those be useful for archi?

I think Mari also has something along those lines.
Yeah we're using triplanar mapping in UE4 for quick previews or for some spline meshes. But it's not as efficient and we're doing huge scenes so we have to optimise quite a bit.
And some objects need mapping anyway like round beams and stuff like that
So I've been testing the waters the past few evenings and I'm amazed at how convenient Houdini feels to use even tho it's so complex. It feels like they really made an effort to make it fun to use.

I'm wondering tho why not more people are adapting the procedural workflows?
>I'm wondering tho why not more people are adapting the procedural workflows?

Well, there are a number of reasons that I can think of...
1. high barrier of entry typically scares people away, especially when Houdini is "another thing to learn", rather than a de facto replacement for any of your existing packages. For example, while you can make amazing and unique things in Zbrush alone, more often than not it's an "organic" supplement to your toolkit, much like Houdini is a procedural supplement.
2. requires abstract thinking to achieve the end result, but most 3D artists are just that - artists, and prefer a "hands on" approach to making things, rather than dissecting the end result into it's competent parts. The artists way is ultimately faster as you can smoke and mirror your way to the end result, but the procedural way would let you make infinite variations after the fact, which admittedly is often not necessary.
3. Not all forms of content creation can suitably translate into a purely parametric workflow, the kinds that are, namely: physics, particles, simulation and destruction, Houdini is obviously already very well known for that. I suppose architecture is also a prime candidate, it's just that they've never really delved into that market.
File: crankit.jpg (18 KB, 700x525)
18 KB
All art requires abstract thinking. Procedural tools like Houdini, are the only tools that let you leverage the full power of the medium, which is computing. If you don't understand computing you won't get procedural tools.

Procedural tools do not just let you make infinite variations after the fact, they let you make infinite variations during the act of creation, aka tweaking, it is a key advantage that is often over looked. Art directors do not often look at something after 1 pass and say good enough.

All things can be parameterized. Some things are more obviously parameterized than others, because understanding them is impossible without paramterization.
I guess the biggest difference, which is something that I have noticed since I started working with actual 3D artists instead of architects, is that CG does not requirequire precision. CG is not going to leave the digital realm so it doesn't matter if something doesn't match the concept art exactly. So I guess it's not that common for 3D artists to make many iterations or have whole parts of a project change completely over and over again.

Another thing is repetition. Architecture is mostly repetition of already designed elements. I guess a 3D artist will strive for creating something unique and new. Architects come up with an idea and then sort of kitbash it together with existing elements. Custom pieces and materials are made but drive up cost exponentially.

So I guess the way of thinking is entirely different, and the demand is different as well. To me, parametric workflows are the only efficient thing when dealing with thousands of instances of repeating geometry that might need to change over and over..

>Art directors do not often look at something after 1 pass and say good enough.
Neither do architects or their clients.
It is a huge advantage to be able to change elements in real time while the boss is looking over your shoulder. I've had so many situations where I made ie a parametric truss and we were able to tweak it in real time until it was exactly what we needed. 50 iterations in 3 minutes.
Sure it takes a bit longer to set up but the end result is so worth it.
3d artists =/= architects

And not all architects = 3d artists

The way many of you self taught blender/max fags talk about architecture like it's just creating something that looks appealing is bonkers. >>572219 is a great example of this. It has no place in a discussion of architecture, it's unrealistic and used only for artistic purposes.

The reason programs like sketchup are widley used within the industry are due to their simplicity to take exactly what architects put in cad files and allow that to be represented in 3d. No one argues skp is competitive with something like max for character creation and game art.
>exactly what architects put in cad files and allow that to be represented in 3d

With rise of real time rendering and using game engines for Archviz this is not enough anymore. CAD Software doesn't care about polycounts or UV mapping or smoothing groups. But those things are relevant now.
I 100% agree those softwares are not enough with how good archvis can look now, but an optinozed house model for a game in unreal is not the same as a real model with all interior structural components that you would model and show in cad programs or skp. Currently there's a big void with nothing bridging this gap from CAD to unreal.

I think twinmotion 2018 looks quite promising, they have upgraded to use unreal.

BIM helps with this, revit can be exported to max or rendered directly with vray
We actually used Twinmotion before switching to UE4. It's pretty decent for architects to quickly make something, but very limited for professional work and also in scale. We mostly do large retail stores and malls and rendering those in real time with good graphics requires the power and optimization options of a proper engine. We will update our license to TM2018 for quick projects tho. But I doubt it will have any of the under-the-hood options available in UE4. But it's def a step in the right direction.
One problem is that in general architects obviously don't know anything about mapping or how to optimise a model for 3D. Like I said earlier, their workflow and software is tailored for irl precision instead of graphics and FPS... Which is also the problem with this:

I have not worked with Revit but the stuff you get out of Archicad is obviously not gameready. So I doubt Revit is different. We will probably see some improvements in that regard in the next years. But unrelated to that, these softwares might be parametric in some way but I wouldn't really call them procedural.
>I have not worked with Revit but the stuff you get out of Archicad is obviously not gameready

hence the direct max export
My company is in the same boat really. But no one here has figured out a good enough workflow to unreal yet without much wasted time taking everything to 3dsmax to make sure the UVs are right, but when doing large urban areas, resorts, big hotels and malls it's a lot of time wasted going back and forth trying to get the light maps right.

Fingers crossed TM 2018 has figured out a way to streamline this.

Currently we are doing many VR presentations with unity and that's going well, with earlier
drafts in twinmotion 2017 for quick fast fly around showcases
Guess there is no other way especially when BIM becomes the standard. But procedural tools can help with this. For example, yesterday we set up a Grasshopper script that finds specific elements in a file and replaces them with their gameready version.
So if we get a plan with 2000 columns, we can just make one column with nice bevels and UVs and it auto places them where they need to be.
Basically we're trying to have gameready versions of all kinds of architectural elements but obviously they must be procedural so they can change with the iterations..

We haven't made it to VR yet. Most clients are upper management types who don't want to spend their time in a HMD and wouldn't know what to do with it anyway. But it's cool to see someone else using TM. I'm guessing the new release will increase their market share substantially, it's looking really nice.

As for mapping large areas, if you're familiar with Grasshopper, look at the human plugin. I'm not going to say too much, but it's possible to set up a script for UV mapping large planar surfaces and have the UV shells match your desired texel density... There's also a custom mapping node that sort of assigns UV points to mesh vertices but we have not gotten that to work properly. That's why I'm looking to Houdini because their UV tools are lightyears ahead of Rhino obviously.

As for roads and such, we just build them in UE4 with spline tools. Beats modeling them imo.
Trying out Houdini right now, and it blew my mind when I realized that all of the program's aspects exist under the hood as operators on the same network. the program seems segmented, but it's really for your sanity, everything from making a cube to rendering the scene is a node and it's all interoperable, so you make even advanced objects like oceans, landscapes, booleans, particles and such interact in ways that directly affect each other and still go back and charge anything at any time. I'm surprised this tool is rarely talked about, it seems fantastic. gonna go back to plowing through tutorials right away.
its rarely talked about here it seems but for any pyro, fluid, rigid/soft body sims and other FX work its basically the king; everybodys using it

a lot of people see modelling in it as a hassle and it is when you're first getting started and only doing really basic stuff without building any frameworks. its just a big bite to take at first
Houdini works as an extension of the computing enviroment in general. A .hip is zipped directory structure, and it can actually be unzipped and operated on.
/3/ in a nutshell
Same. I just won a year of it in a contest and I'm trying to absorb as much tutorial content as possible but good ones are kind of hard to come by.

Also, a lot of these tuts use Redshift or Octane but I have AMD. Any suitable openCL replacement for these or do I just have to wait?
>I'm wondering tho why not more people are adapting the procedural workflows?
Blender Foundation has talked about wanting to "make everything nodes", basically kind of like Houdini. I think that got side tracked since the focus now is on updating Blender to have the basic modern amenities that commercial packages have, like support for a proper PBR based workflow.

The other 3D packages (that aren't for Archviz or CAD) there's not really a demand, they don't have the flexibility to do something crazy, and if you try to take a procedural approach to 3D modeling (i.e. heavy use of Modifiers, procedural texturing / Materials, procedural UVs) you'll start to run into problems, not the least of which is the significantly increased memory burden since those softwares weren't designed for that. Whereas Houdini likely does various tricks to optimize / minimize the impact of a gigantic modifier stack / node tree type organization.

The TL;DR is Houdini isn't the industry standard, using it requires a completely different way of doing things, and there's not much incentive for Autodesk to do things differently.
The BF didn't get side tracked, they realized that before they can even start with the 'everything nodes' project they have to fix the core and UI first.

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.