AI rendering has moved from a novelty to a core part of the architectural design process, and Chaos Veras is leading that shift. With the arrival of its new "Nano Banana" rendering engine, Veras can turn a Revit, SketchUp, Rhino, or Archicad model into a photorealistic image in seconds, guided almost entirely by the words you type.
That is exactly why prompting has become a skill in its own right. The better you can describe light, material, geometry, and mood in plain language, the more control you have over the final image, and the fewer iterations you need to get there. Below, we break down how the Nano Banana engine works, the prompting framework Chaos recommends, and the new workflows it unlocks for architects, interior designers, and urban planners.
Nano Banana is the AI engine now powering Chaos Veras, and it represents a meaningful jump forward for AI-driven visualization in three ways:
In Veras, your prompt is no longer just a creative brief — it is your complete set of controls. The most predictable results come from treating a prompt like a formal specification rather than a single sentence. Chaos recommends structuring prompts into five sections:
Pro tip: the Constraints section is what replaces the old override sliders. This is where you tell Veras exactly how strictly to follow your original model, and it is the single biggest lever for predictable, on-brief output.
Long, unstructured paragraphs can work, but breaking your prompt into the five sections above gives the AI a much clearer brief, and gives you a repeatable template you can reuse across projects.
Use the Constraints section to explicitly define what should never change, such as massing, window placement, or facade rhythm, versus what you want Veras to reimagine, like a material or façade pattern.
Fight the "AI look" by naming specific material imperfections: visible wood grain, concrete pores, subtle water stains. Explicitly asking Veras to avoid an over-smoothed, plastic finish noticeably improves texture fidelity.
Camera-specific language, such as lens focal length, angle, and framing, gives you the same level of compositional control a photographer has on site, without touching a single slider.
Beyond single-image prompting, the new engine enables several workflows that meaningfully speed up early-stage design and client communication:
For firms already working across the Autodesk and Accruent ecosystems, this shift means AI rendering is no longer a separate, specialist skill; it is an extension of the modelling work your team is already doing in Revit, Rhino, or SketchUp. Learning to prompt well is quickly becoming as valuable as knowing your modelling software.
Nano Banana is the new AI engine powering Chaos Veras. It improves fidelity and realism when converting 3D models into photorealistic images, unlocks new workflows like 2D-to-3D conversion and mood-board-driven rendering, and replaces override sliders with a more direct, prompt-based conversational experience.
Veras works with Revit, SketchUp, Rhino, and Archicad models, turning them into photorealistic images guided primarily by text prompts.
Chaos recommends structuring prompts into five sections: Task (the high-level goal), Style & Art Direction (aesthetic references like a publication or photographer), Components (subject, setting, and micro-details), Lighting (photographic language like "golden hour" or "volumetric rays"), and Constraints (what must stay true to the source model and what to avoid).
The Constraints section replaces the old override sliders. It tells Veras exactly how strictly to follow the original model — specifying what must stay true (such as massing, window rhythm, or vertical lines) and what to avoid (like a sterile, empty scene or low-quality entourage) — making it the biggest lever for predictable, on-brief output.
Name specific material imperfections in your prompt, such as visible wood grain, concrete pores, or subtle water stains, and explicitly ask Veras to avoid an over-smoothed, plastic finish. Using camera-specific language — lens focal length, angle, and framing — also gives you photographer-level compositional control.
It enables 2D-to-3D conversion, mood-board-to-reality interior design, diagram creation (exploded axonometrics and analytical diagrams), using a reference image as input, design persistency across camera angles, furniture placement from reference images, and intelligent BIM integration through the Enscape bridge with Revit, Rhino, SketchUp, Archicad, Vectorworks, and Autodesk Forma.
For firms working in the Autodesk and Accruent ecosystems, AI rendering is becoming an extension of modelling work already being done in Revit, Rhino, or SketchUp — making prompting skill as valuable as modelling skill itself.
This article covers the core framework, but the full "Prompt Like a Pro with Chaos Veras" guide includes the complete prompt templates, a library of style, lighting, and mood keywords, an artist and design-era reference table, and worked before/after examples for every workflow above.