Nashville, Tennessee — At Autodesk University 2025, visualization software leader Chaos made a strong case that AI no longer optional — it is critical for survival. From its roots in rendering engines like V-Ray, Corona, and Enscape, the company is now positioning itself as an architect of AI-driven visualization and design workflows.
“We are past the hype,” said Bill Allen, who joined Chaos when it acquired EvolveLAB earlier this year. “AI tools are already being used by your competitors to automate construction documents, to win proposals, to accelerate iteration. If you don’t adopt them, you’ll fall behind.”
From Marvel to Hospitals: A Cross-Industry Visualization Application
Chaos describes itself as a “world leader in visualization.” Indeed, its software powers are mighty, making eye-popping graphics for architectural design reviews, construction coordination… and even Hollywood blockbusters.
“When you see a Marvel movie or Star Wars, odds are that’s our software,” said Roderick Bates, Head of Product Operations at Chaos. The cross-pollination between media and entertainment (M&E) and architecture, engineering, and construction (AEC) has become one of Chaos’ key differentiators.
Massive file-handling techniques developed for architecture are now being used in film. Conversely, cinematic rendering and lighting approaches are being adapted into real-time AEC visualization. That convergence has shaped Chaos’ strategy: to build an ecosystem that works seamlessly across disciplines and integrates AI not as a gimmick, but as a productivity engine.
Surveying the State of AI Adoption
For proof, Chaos cites its industry surveys. In Spring of 2025, 56% of design firms reported actively using AI in workflows, and another 24% were in pilot or planning stages. In other words, four out of five firms were already experimenting with AI.
The growth has followed the ups and downs of the Gartner Hype Cycle. At Autodesk University 2022, only four sessions touched on AI. By 2023, that number jumped above 50, as the hype cycle peaked. In 2024, the number dipped as unproven ideas fizzled. Now, in 2025, Chaos argued, the industry has entered a maturity phase — tools are real, results are measurable, and ROI is tangible.
Ethics and AI Governance
Roderick emphasized their commitment to responsible AI. “We want to elevate creativity, automate the mundane, and innovate without compromise,” Allen said. “That means ethical sourcing of data, transparent disclosure of models, and clear governance.”
Key principles include:
- No training on customer data. Proprietary design files remain off-limits.
- Transparency at the feature level. Each AI feature (enhancer, generator, upscaler) may use a different engine, and Chaos documents which.
- Up-to-date disclosures. “If an AI feature cites a copyright from 2024, that’s a red flag,” one executive warned.
- Plain-language policies. “Our lawyers told us: no 80-page EULAs. Firms want to know in simple terms what’s happening with their data.”
The company even engages directly with customers’ legal and IT departments to validate compliance. “Good enough is not enough,” Bill Allen said. “You must know exactly what’s happening with your data.”
What? Video from Images?
One of the presentation’s highlights was Veras, the AI rendering tool integrated with Autodesk Forma, Revit, Rhino, and SketchUp. Unlike generic text-to-image systems, Veras leverages actual project geometry to generate rapid iterations.
“You render a white-box model, set a background, and every 20 seconds you can test a new facade or material,” Allen explained. Clients can then compare side-by-side options in tools like Miro or PowerPoint.
A newly released Veras 3.0 Image-to-Video feature lets users animate still renderings (watch it here). Prompts like “pan left,” “add sunset,” or “people walking” generate cinematic clips. Chaos demonstrated uses ranging from architectural flyovers to playful minions wandering through plazas. More experimental workflows include form-finding — using AI to morph building geometry on the fly.
Enscape, Chaos’ real-time rendering engine, is also being retooled with AI. Its AI Enhancer can refine people and trees in a scene, alter age or ethnicity to match project context, or generally “make the rendering pop” without having to render in another package.
“Think of it as taking something that’s already accurate and turning it up to 11,” Allen said.
Material Intelligence: From Photos to Libraries
Another pain point Chaos addressed is material creation. Traditionally, designers spend hours scanning materials into libraries using tools like Adobe Substance. Now, Chaos’ AI can generate seamless, high-quality materials from a photo, automatically correcting lighting.
The roadmap includes an AI Recommender, which suggests materials based on mood or visual input. “It turns a dead-end image into the start of a real design,” a speaker said. This feature is expected in 2026.
“We used to model to render. Now we render to the model.”
Control Layers and Iteration
A recurring theme was control. Designers often complain that AI changes too much or strays from intent. Chaos’ tools introduce control layers:
- Selective overrides: change only fins on a facade, or recolor floors, without affecting the rest of the building.
- In-painting with constraints: adjust glass ratios for hot climates, add branding colors, or replace wood with steel.
- Iteration loops: present to clients, receive feedback, re-prompt, and update — all in minutes.
The goal is to keep architects “in the driver’s seat.” “AI should feel like an intern you direct, not a boss who overrides you,” Allen said.
Efficiency in Action: From Lunch to Presentation
To underscore efficiency, Chaos pointed to examples created “during lunch” at AU 2025. Using image-to-video, staff generated interior scenes with lighting changes and animated furniture. By afternoon, the clips were embedded in the live presentation.
“That turnaround used to take days with traditional tools,” said Roderick. “Now it’s hours — sometimes minutes. That’s the real efficiency AI delivers.”
The Night of Doritos
While flashy renderings draw attention, Chaos also targeted the grind of construction documentation. The company’s Glyph tool, an AI-powered Revit add-in, automates repetitive tasks:
- Generating plans and elevations by level.
- Tagging rooms, walls, and doors with collision-free dimensions.
- Creating and populating sheets, even spilling overflow views onto new sheets automatically.
- Bundling tasks into macros that run entire documentation sequences.
Bill recalls when he was with an architecture firm and was rushing to get a drawing set out.
“It was about 11:30 at night. I was working with my architect. I said, ‘Hey, man, I’m gonna go home, take a shower, get some rest. I’ll help you in the morning. We’ll hit print, get this thing out the door.”
“He’s like, ‘Yeah, that’s fine.’”
“I went home, took my shower, came in the next day, he was sitting in his same chair, he had his same shirt on, he had a bag of Doritos on his desk, and there was this orange handprint on his shoulder.”
“You didn’t go home last night, did you?”
“He had pulled an all-nighter trying to get the construction drawing set out the door.”
“Glyph aims to eliminate that,” Allen said.
Glyph includes a Copilot interface that allows users to issue natural-language prompts, bringing ChatGPT-style interaction into Revit.
Upscaling, Enhancement, and Targeted Edits
Chaos also showcased AI upscaling, boosting 2K images to 4K without having to rerender. It’s just like on TV or movies when a person says, “enhance this image,” and license plates and fuzzy faces are suddenly clear.
“In 2025, that’s real,” said Roderick.
Targeted enhancement was another standout: updating a single asset in a scene, such as replacing stock young people with seniors for a retirement community. The result, Chaos claims, is photoreal quality without re-rendering or custom asset creation.
Client Resistance and Legal Risks
Not all can be expected to be enthusiastic about AI tools. Some companies have made “no AI” policies, often for legal reasons. That can force firms to disable AI features entirely on specific projects. Chaos is exploring “switchable instances” of software — one with AI, one without.
The bigger challenge may be fees. Executives suggested firms may soon offer two price structures: one with AI acceleration, one without. “When clients see the cost differential, they may change their tune,” Roderick predicted.
Looking Ahead: No Buttons, Just Prompts
Asked about the five-year outlook, Allen cited a conversation with Sam Altman of OpenAI, who predicted AI would disrupt creative work first, then cognitive labor, then physical labor.
“We all thought design was immune. Then AI started generating buildings from prompts,” Allen said. Looking forward, he sees agentic AI — fleets of task-driven agents — augmenting workflows. “You’ll still be the orchestrator, but you’ll direct more throughput than ever before.”
Broderick offered a more radical vision: “In five years, my renderer may not have any buttons at all.”
Conclusion
Chaos’ AU 2025 message was clear: AI is mission-critical. With Enscape Premium, Veras 3.0, Glyph automation, and a suite of AI enhancers, the company is betting that firms will embrace a new workflow paradigm — not to replace architects, but to liberate them from drudgery and accelerate creativity.
“Choose wisely,” says Roderick. “Not all AI is equal. But the firms that wait will find themselves playing catch-up.”