The archviz studio Polymachine team talked about their field, its current challenges, the role of Houdini and procedural approach in their pipeline, and more.
80lv: Could you tell us a little bit about your studio? Where are you guys based, what projects have you worked on? How did this whole journey start for you?
We’re based in Zagreb, Croatia and we specialize in architectural visualization. However we like to dabble in other areas of CG as well because sometimes archviz can get a bit repetitive, and it’s always good to broaden one’s skill set. Although some of us have been in the archviz industry since the early 2000s, the roots of Polymachine started with Dean Niskota and Hrvoje Čop in what was Studio Niskota back in 2014. After doing really well in the local Croatian market, in 2018 we were joined by Martin Ferić, a long time business partner of ours, and decided to regroup, rebrand and take the next step in our creative journey. We’ve gone from three people to a team of six people, moved to a new office, and we’re still growing. It’s been quite a ride, to say the least.
Throughout the years we’ve worked on hundreds of architectural projects – interiors, exteriors, animation, VR and so on. We’ve also been working a lot with furniture manufacturers, doing fully CG catalogs for print and product placement. The projects we like most, however, are ones which really challenge our skills. We don’t like staying in our comfort zone for too long. This approach led us to have a really good relationship with software companies because we strive to push their software to its limits. And well, sometimes break it. But staying on the bleeding edge of technology has definitely always been one of our focal points.
Archviz Industry Nowadays
80lv: What are the most challenging things that archviz industry currently faces in terms of production? What are the challenges that increase your production time, what are the problems that you come across?
From an industry perspective, we’re all still very much succumbed to Blinn’s law, i.e. the idea that as computing power increases, so does the desired quality of content keeping render times mostly unchanged throughout the time. However, client demands are increasing rapidly, and combined with the fact that creative industries are constantly shifting towards ever quicker turnarounds, this poses a problem of keeping up with those client demands. Due to their complexity, architectural projects are especially difficult to produce both fast and with great quality, so some artists compensate for this fact by doing a lot of post work which doesn’t rely that much on render power or extremely complex 3D data, while others, like us, work towards optimizing their 3D workflows and having a lot of machines and specialized software to do the heavy lifting.
From our point of view, we’re lucky to be living in times where the tools we use advance so fast and consistently enable us to do more in less time. But if we had to cherry pick a single most influential factor to production time, it would be the project changes during production. As anybody working in CG knows, once you’ve been in production with a project for a while, going back to change stuff can set you back a lot, and it usually ends up being easier to just redo those parts from scratch. This is one of the main reasons we stick to 3D since we believe that it facilitates easier changes than doing post work, which is essentially wasted if something significant changes, or even something small like a camera angle. But, it needs to be said the whole more postless post-debate is a matter of personal preference though, we’re not claiming our way is the universally right way, but it works for us.
Introducing Houdini into the Pipeline
80lv: Why did you start using Houdini in your production? This is by far not the most popular tool in your industry and most people just stick with 3ds Max. What were the things that you found most useful and how did you start applying the new techniques?
Houdini came to be an experimental part of the pipeline since we’re always on the lookout to be more efficient and find better solutions to everyday problems. And yes, figuring out how and what it actually does is no small feat. Honestly, we’re just cracking the surface of what Houdini can do, but it already solved a few problems we’ve been having for a long time. It’s more data management than what you would think of when you say ‘3D graphics’, but that’s exactly where its power lies. Houdini is all about ideas and procedures, not about the actual geometry you’re working on. If you get it right in principle, whatever you made will work in a lot of different scenarios. Suddenly things that seemed complicated to us in the past, such as doing volumetric clouds or fog, fracturing walls etc. became a lot more doable. Once you get to that point, it practically calls on you to investigate further and see what more can be done.
Applying Procedural Approach
80lv: Could you talk a little bit about the way you’ve been working with the procedural nature of Houdini in terms of asset generation? How did you use its capabilities and what useful lessons have you learned during the process?
Not everything can be solved by the procedural approach. The problem with architecture is that you are working on something that is already designed in a specific way. If you’re making a game level, you might generate a bunch of procedural assets and your level design will have a look and style which is a result of how your assets are made to work. You don’t really get that freedom with architecture – i.e. you are much more constrained with the design you’re trying to recreate. Where we see Houdini helping is not necessarily with the architecture itself, rather with everything else that comes with it. Imagine having to place glasses and plates on a hundred restaurant tables or filling shelves with books. There are solutions to this already, but we feel Houdini will help us drastically simplify and speed up things like this.
80lv: Another interesting use of the procedural approach is adding more variation to break the perfect look of some elements. You’ve mentioned it in your recent tutorial. How could Houdini help you here? What are the advantages of using this tool for randomization? We naively thought that archviz was a more precise field, where everything should look as realistic as possible.
Well, the reality is imperfect. Archviz, on the other hand, is a fine balance between clean – which is what clients usually aim towards, and ‘somewhat dirty’ – which is more realistic. After all, the 3D imagery was notorious for decades for being too perfect, but now we’re at the point where all those subtle imperfections are much easier to produce than ever before.
Randomization is something we always worked a lot on and tried to streamline because humans are really bad randomizers by nature. We’ve developed quite an arsenal of tricks for it to date, and coupled with a few in-house tools, we can randomize objects and materials relatively quickly. However, due to its procedural nature, Houdini is a perfect tool for this, and we’re looking into it for things like stone walls, roads, window blinds. Things which are usually a lot of work, or tricky to do manually. Having a setup for any of these things in Houdini means you can reuse it to your heart’s content.
80lv: How do you feel Houdini copes with the production of reusable assets? It would be awesome if you could show some examples of how you can quickly generate versatile architectural elements without having to model and assemble everything by hand.
Our first real experiment with Houdini Digital Assets was a window generator. Windows can be a real mess in some projects because of a large variety of sizes and types. We’ve been looking for a decent solution to this for years, and we’re on the verge of having one finally. Right now we can just use the topology of simple planes to define the shape of the windows and pick spline profiles for getting the exact look we’re after. It’s fast, simple and thanks to the fact the asset appears as a modifier in 3dsmax’s modifier stack – very easily modifiable.
Right now, we’re working on adding animatable parameters to this asset, as well as other effects like integrated blinds and more. Once we solve this, hopefully, windows will also go to the ‘I’m not doing this by hand’ list. And we really like adding things to that list.
80lv: How do you work with materials in your projects? We wonder if you stick with some scanned stuff from Megascans or if you use SD. How procedural is your texture work and what are the biggest challenges you have with this type of content?
We use a lot of textures from stock sites like Poliigon, some Megascans, and a lot of Substance Designer. We jumped on the Substance train several years ago, back when people didn’t really see the appeal of it for archviz. We’ve found that for any kind of pattern-based material (like textiles, tiles, wallpapers) Substance is massively faster than doing things in Photoshop or even looking for decent textures online. Most often, manufacturers don’t provide textures which are nearly good enough, so recreating them in Substance ensures you have sufficient quality and absolute control over the result. On the material side, we try to keep things simple, but it’s always challenging to texture huge surfaces while keeping the textures non-repetitive, so a lot of times we end up with pretty huge shader trees, mixing several maps and masks to make materials which just work.
Currently, we’re experimenting with Open Shading Language or OSL, which was recently added to 3ds Max. It opens up a whole new world of possibilities and is renderer-agnostic by design. That’s quite important to us since we render in both Vray and Corona. Unfortunately, shader writing is surprisingly hard to get into, since you have to write code which works for any arbitrary point in 3D or UV space, but we’re slowly getting there.
Does Archviz Need Simulations?
80lv: It would be awesome if you shared some thoughts on how Houdini simulations like clouds, water, etc. could help you in your work. Do you want to use those features? Isn’t it overkill for archviz?
We’re really excited about simulation. We’ve used Phoenix FD on several projects so far because however you choose to fake some things, they will always look better simulated. Houdini offers a really powerful simulation toolset, and we’re looking forward to using more of it. Clouds are definitely on the list, as well as fog, fire, and water. The obvious advantage to simulation is that it can interact with the scene in subtle ways. Clouds can be in real scale and cast shadows on the scene, fire can light a space in a realistic way.
It might be overkill for some, but we like it, and we like the control we get with it. For us, one of the worst parts of this job is going online and spending hours searching for backplates that fit what we want to achieve or textures which are good enough etc. Having a pipeline where we can just create whatever we envision is where we want to ultimately end up. Besides, if you love 3D as much as we do, having an excuse to use a ton of different tools and experiment with stuff is the most fun you can possibly have.