How Chaos accelerates 3D visualization workflows with AI

Keith Shaw: Hi everybody. Welcome to DEMO, the show where companies come in and show us their latest products and platforms. Today I'm joined by Roderick Bates. He is the Head of Product Operations at Chaos. Welcome to the show, Roderick.

Roderick Bates: Hey, thank you. Keith: Chaos.

What a cool name for a company.

Roderick: It really is. It's a fun name. You always sit next to people on airplanes, trains, whatever, and they ask, “Where do you work?” I say, “Chaos.” That's pretty nice. Keith: Yeah.

For people who might be unfamiliar with Chaos, tell me what the company does and what you're going to be showing us here today. Roderick: Absolutely.

What Chaos does is we make visualization software. That means we can take whatever you might have modeled in a computer program - pretty much any program - and render it.

We allow you to apply materials, lights, scenes, and assets and basically make it look photorealistic from that computer model.

Keith: And today you're going to be showing me some AI enhancements. Roderick: Absolutely.

It's a workflow we've been building for almost 30 years. When it started, obviously there was no AI. It was a very manual process. We had a lot of tools that evolved from requiring you to be an artist to allowing you to do things more quickly.

Now with AI, it's a whole different game. We're leveraging all of our years of experience to make AI as controllable and applicable as possible for specific industries. I'll be showing some examples that are specific to architectural design.

Keith: Obviously with what Chaos can do, there are a lot of different industries. Who within those industries is usually the main user? Is it designers, architects, or others? Roderick: Right.

We work in movies, advertising, and video games. But in architecture we're primarily seeing designers using it. That includes architects themselves or interns working with them. What's fascinating with AI is that it provides a new level of accessibility. It's becoming a design communication tool between clients and architects.

Instead of a client saying, “The door should be a little over this way,” they can actually use the AI tools to adjust things visually and say, “No, that's what I want.” So they can communicate visually instead of purely verbally.

Keith: One challenge a lot of people have is simply how long it takes to do something. The AI enhancements help speed that up. Roderick: Exactly.

One problem was how long it took. Some tools replicate physics and light, and that could take 20 minutes for a single image. If you're making a movie at 30 frames per second, you're talking about serious time.

Other tools introduced real-time ray tracing, so you could replicate physics and materials in real time. But you still had to set up the scene. That wasn't extremely slow, but it still took time.

Roderick: If you're early in a design process and maybe you're not even getting paid yet, you might just need to produce a concept to excite a client. That's where AI really shines.

Or maybe you show a client something you spent a lot of time on and they say, “Love it, but can you take the fruit off the table?” Then you have to go back and redo it.

With AI you can do targeted enhancements, select one thing, and say, “Change it this way.” You can create a hybrid workflow between traditional tools and AI to dial in your intent without starting over.

Keith: Was a lot of this based on generative AI from recent years, or have you been working with more traditional forms of AI before that?

Roderick: That's the funny thing about AI. There were many things we did that could technically be called AI. Maybe they were more deterministic in the sense that a given input produces a clear output.

Now we're seeing systems where you provide very little input and the AI generates something based on its training and imagination. We like to use both. We focus on workflows where we can drastically accelerate very specific tasks.

We're starting to see that not only in full images but also smaller elements, like materials. For example, in architecture you might want a specific flooring material. Previously you'd take a photo, bring it into Adobe software, create a digital material, and apply it to the scene.

That could take an hour or more. AI can now take that picture, create the material, and apply it to the model in about five minutes. Keith: All right.

We're talking about visual work, so let's get to the demo. Seeing is believing. Roderick: Absolutely.

We're launching the program in the context of Rhinoceros, which is one of the many architectural programs we integrate with. You can see we're launching Enscape directly within the program. What we're doing right now is reading all of the model content and creating the scene.

This model started as a very basic setup. To access our AI features, they're embedded directly in the tool. On the right-hand side you'll see preset prompts. You can always enter your own, but our visual artists created prompts designed to get the most out of AI.

Roderick: The system reads the metadata we extract from the host application and uses that to generate something faithful to the original geometry.

Keith: So it's the same building, but now it's in the summertime. Roderick: Exactly.

Now we can take that image and adjust it. For example, right now we're in Boston and it's cold with snow on the ground. So let's create a snowy scene with icy trees and bundled-up people.

Traditionally you'd go back to the modeling program and add all those weather effects manually. Here it's dramatically faster.

Roderick: There's our scene. It looks pretty good. But people often want animation. It doesn't look quite right if it's not actually snowing. So we'll take the same view and generate a short video clip. On the right side you'll see the reference image we're using as the starting point.

We'll add snow and minimal camera movement so the scene feels alive.

Roderick: As the AI processes, it generates a video. Obviously it takes some computation time, but compared to traditional animation workflows where you'd model everything and plan each movement, this is remarkable. You can create several clips like this and stitch them together into a narrative.

When clients ask what their building will look like, showing them something like this makes it clear. Architectural plans alone don't always communicate the idea.

Roderick: Now let's look at interior examples. One challenge is creating accurate materials. You don't want to show a material that won't be used in the real project. Within our material editor we have an AI material generator.

You upload a photo, it scans the image, adjusts lighting, determines the correct crop, and generates a full digital material with proper reflectivity and roughness.

Roderick: Traditionally that workflow could take 20 to 45 minutes per material. Here it takes just a couple of minutes, which adds up significantly over a day or a week.

Keith: But that doesn't look right. Roderick: Exactly.

That's where hybrid workflows come in. AI gets you most of the way there, but you still use manual controls and the artist's eye to refine the scale and details. We think of AI as a “better brush.” It's augmenting the artist rather than replacing them.

Roderick: Let's try another example with a concrete wall treatment. Once we generate it, it quickly populates the scene. From there you can adjust the scale and details. The AI material system generates all the necessary properties so it behaves like a real-world material.

Keith: I like this part here too.

Roderick: This part is wild.

Keith: It looks like GTA 4.

Roderick: That's because it's real-time software. As you move through the scene you see everything immediately. But real-time scenes use simplified geometry, which isn't great when you zoom in.

So when we enhance these characters using AI, it processes them in the cloud and uploads the improved version back into the scene.

Keith: That obviously looks better.

Roderick: It's like putting on glasses. Suddenly everything is sharper - hair, hands, and facial details. But now we notice the guy on the right looks unhappy. He's watching a movie, so he should be smiling. Since our system recognizes people and facial expressions, we can apply targeted AI adjustments.

Roderick: Using a mask, we select that person and tell the system to change the expression. We could also make him younger, older, or remove facial hair.

Keith: If you had to go back to the modeling program to do that, it would take a long time. Roderick: Exactly.

Here it's real-time.

Keith: That's a lot of time saved.

Roderick: And there he is. Now he's smiling. It's a quick adjustment that makes the scene feel more natural.

Keith: But again, this is something a human designer would recognize. You wouldn't just randomly say “make this guy happier.” Roderick: Right.

If you did, you might not get the result you want. That's why we think of AI as a design companion. We have decades of experience in this field, and our users have strong aesthetic instincts.

Our goal is to balance AI automation with traditional workflows to get the outcome they want.

Keith: How can customers try this? Do they already need to be a Chaos customer?

Roderick: Not at all. We offer a free trial at chaos.com. You can explore all of our software there. The trial lasts about 30 days, so you have plenty of time to experiment. What's great about the AI workflows is that you don't need to be an expert.

If you've always been intimidated by this type of software, you can jump in, have fun, and create something beautiful. Keith: All right.

Roderick Bates from Chaos. Thanks for coming in and thanks for the demo. Roderick: My pleasure.

Keith: That's going to do it for this week's show. Be sure to like the video, subscribe to the channel, and add your thoughts below. Join us every week for new episodes of DEMO. I'm Keith Shaw. Thanks for watching.