How to render with AI without leaving Revit or SketchUp: a 2026 workflow guide

Key takeaways:

• Every external AI tool iteration carries a hidden round trip cost. Export, upload, prompt, reconcile. Running Veras natively inside Revit or SketchUp eliminates that tax and keeps the iterative rhythm intact.
• Your prompt controls both the visual style and how closely the AI follows your existing geometry. Give it as much or as little creative freedom as your design stage requires.
• Smart Selection means you can change one detail without regenerating the full render. Target any region, object, or material with a single click, choose Modify or Replace, and re-render just that area.
• Veras fits into your existing pipeline rather than replacing it. Use it for concept and DD work, then hand off to Enscape, V-Ray, or Corona for final deliverables.

AI-powered visualization has become a standard part of the design development process. With the majority of architecture firms now reaching for an AI tool at some point between schematic design and client presentations, whether to quickly test a material palette, communicate massing intent, or produce a mood image for a stakeholder meeting. The challenge is that most of those tools live outside the modeling environment, which means every iteration carries an export-upload-reconcile tax that quietly erodes studio time.

Veras addresses this directly by running as a native plugin inside Revit and SketchUp. It reads your active 3D view—geometry, camera, and materials—and generates AI imagery in seconds without requiring you to leave your modeling environment or transfer a single file.

This guide walks through how to use Veras inside Revit and SketchUp in 2026: what to click, how to write prompts that produce consistent results, and which controls keep every render anchored to your actual geometry.

Table of contents

→ The high cost of the export-upload-prompt cycle
→ How to set up your first AI render in Revit
→ Iterating conceptual designs within SketchUp
→ Mastering the Veras feature set for precision control
→ Comparing integrated AI workflows to standalone rendering tools
→ Three habits for BIM managers adopting Veras
→ Integrated vs external AI rendering for Revit and SketchUp
→ Frequently Asked Questions

The high cost of the export-upload-prompt cycle

Every minute spent moving a model from Revit or SketchUp into a browser is a minute not spent designing. The cost is rarely a single bottleneck. It is dozens of small interruptions that fragment a working session and pull the design lead away from the building itself.

Treat the friction as a stack of real, measurable costs:

• Lost studio time. Export, file conversion, upload, prompt, download, reconcile. Repeat for every iteration. A 30-second AI render can sit inside a 10-minute round trip.
• Lost design momentum. Context switching kills the iterative rhythm. By the time you are back in the model, the idea you wanted to test has cooled.
• File management overhead. Versioned exports pile up. So do half-prompted variants. Without a strict naming convention, your desktop becomes a graveyard of massing_v07_final_REALfinal.png.
• Geometry drift. External tools see pixels, not BIM data. The pattern is familiar: you prompt, get a styled result, then try to reconcile that image with the 3D model still sitting in your CAD app. Proportions drift, massing changes between views, and the AI freestyles on geometry it cannot see.
• IP and data exposure. Every screenshot pushed to a public AI platform is a piece of an unbuilt project leaving your firewall. For competition work, NDAs, and government clients, that is a real concern.

This is why an AI plugin that reads directly from your active view matters in 2026. Veras runs as a native add-in inside Revit, SketchUp, Rhino, Vectorworks, and Archicad using your existing 3D geometry as the substrate, so the imagery stays a valid representation of the model.

How to set up your first AI render in Revit

The Revit workflow is built around your active 3D view. Veras grabs it directly from the ribbon, so there is no FBX, no OBJ, and no screenshot in between. Walk through it the first time on a simple massing study, not a 500MB federated model.

1. Open a 3D camera view: Veras needs spatial depth, not a plan or section—perspective information is required to produce a convincing result. Frame the shot exactly as you want it composed.
2. Launch the Veras panel: Go to the Chaos tab in the ribbon, press the Veras Start button. The Veras panel docks alongside your model and captures the current view automatically.
3. Choose a Veras Preset: In the Explore tab, pick a Veras preset that matches your scene, then click Render or go to the Compose tab for more granular control.
4. Write a precise prompt: In the Compose tab, further customize your rendering. Use architectural terminology, not adjectives. "Board-formed concrete plinth, anodized bronze mullions, late-afternoon sun from the southwest, low contrast editorial photography" beats "make it look nice." Veras treats prompts as instructions with localized overrides, so naming specific assemblies pays off.
5. Set prompt-based geometry control: With render engine V7, geometry override is driven by your prompt language, not a slider. Add phrases like "strictly preserve massing and fenestration pattern" for DD-phase fidelity, or "feel free to reinterpret the roofline" when you want the AI to push the form. If you're on an older engine, use the Geometry Override slider in the Compose tab instead.
6. Render: Choose which render engine you want Veras to use, the aspect ratio, resolution, and how many renderings it should produce, and then click Render. Cloud processing returns imagery in seconds for Stable Diffusion and roughly 10 to 30 seconds for Nano Banana Pro, depending on resolution.
7. Lock your composition with Reference Images: When a render lands that you want to iterate on, feed it back in as a Reference Image. This keeps composition, atmosphere, and lens consistent as you change one prompt variable at a time, such as material, sky, or lighting, without losing the rest of the image.

Pro tip: Save your strongest prompt as a project-level text snippet. Studios that produce 20+ variations a day standardize prompt blocks (lighting, lens, materials, mood) the same way they standardize view templates.

→ Ready to animate using Veras? How to animate your construction with Veras and Revit

Iterating conceptual designs within SketchUp

SketchUp's strength is fast massing. Veras meets it there, turning a rough volume into a styled visualization without leaving the modeling window. The plugin runs as a SketchUp extension, accessible from the Extensions menu or the Veras toolbar after install.

A clean iteration loop for SketchUp:

1. Stage the scene. Set a SketchUp scene with the camera, shadows, and style you want as the starting point. Veras reads the active viewport, so what you see is what gets rendered.
2. Tag intentionally. Use SketchUp tags and components to separate massing from context. The AI infers material logic from the model's structure, so a clean tag tree produces cleaner material interpretation.
3. Open Veras from the toolbar. Launch the panel, confirm the viewport capture, and use the Explore tab to choose a preset, and click on the Compose tab to write your prompt and set your render engine.
4. Iterate without closing the model. If a window proportion looks wrong, fix it in SketchUp, refresh the viewport, re-render. The model is the source of truth, not the image.
5. Use Smart Selection for local edits: Target a region, object, or material with a single click and type a new prompt for that area ("replace planters with mature olive trees"). Choose Modify to preserve geometry while changing materials, or Replace to reimagine the selection entirely, then re-render just that area. The rest of the composition stays intact.

This is where Veras earns its keep during schematic and early DD work: the geometry edit happens in SketchUp, the visual feedback happens in Veras, and neither one waits on a file transfer.

Mastering the Veras feature set for precision control

Once the basic loop is muscle memory, the controls that separate a fast render from a presentation-quality one are prompt structure, the use of reference images, and selection-based edits. Each does a different job.

Geometry control

Veras lets you control how closely the AI sticks to your model's geometry, and how you do that depends on which render engine you're using.

Using Veras Version 7, this is handled through your prompt. Write “preserve the existing geometry” to lock the form down, or use a more abstract prompt like “a brutalist concrete facade" to give it more freedom and creative liberty. Veras understands geometry, material, and style as separate instructions, so you can mix them: "preserve the walls and roof, but turn the facade into Japanese-style charred timber" works exactly as written, which is great for DD-stage studies.

Using Veras Version 6 or below, you’ll find a dedicated Geometry Override slider in the Compose tab that does the same job. Slide it low for strict adherence, or high for creative freedom (perfect for early concept exploration).

Reference Images

Reference Images make it easy to produce coherent iteration sets. Instead of hoping each new render lands in the same compositional territory, feed your previous rendering back in as a reference and the AI uses it to stay consistent, same lens, same atmosphere, same overall feel. Change one variable in the prompt (glass tint, paving, sky condition) and keep everything else anchored to the reference. Reference Image as input is the mechanism behind producing a coherent set of facade studies for a client deck, and it's more reliable than any number-based lock ever was. Here, architect Kate Vera dives deep into the Reference Image feature:

The Selection tool and render selection modes

Smart Selection is the underrated feature. Instead of regenerating an entire perspective to test a single canopy material, use Smart Selection to target it with a single click — by category, object, material, or a custom prompt — and re-render that region only using the new Modify or Replace modes. The rest of the image stays pixel-identical. This protects your approved composition while you iterate on details, and recent selection prompts are saved with thumbnails so you can reuse them across renders without retyping.

Pro tip: Hold Alt to select every instance of a material or object class in a single click.

Image-to-Video and sky replacement

• Image-to-Video. Veras turns any still into a 5-second animation using 12 video presets, useful for adding weather, vehicles, or time-of-day shifts to a client presentation without a dedicated VFX pass.
• Sky Replacement: Swap the sky in any rendered output by typing a sky condition into your prompt, e.g., overcast, golden hour, dramatic storm, without re-rendering the rest of the composition. Useful for quickly producing multiple time-of-day variants of the same view for a client deck.

Comparing integrated AI workflows to standalone rendering tools

External AI image generators and web-based AI rendering platforms can produce striking imagery, but the trade-off is the shape of the workflow. Tools that sit outside Revit and SketchUp ask you to export, upload, prompt, and reconcile. Tools that sit inside your modeler read the geometry directly.

For design development work, native integration is arguably the most important factor to evaluate. External tools strip away materials, lighting, views, and geometry relationships. The AI never sees your BIM data, only a screenshot of it. Veras runs as a plugin inside Revit and SketchUp (and Rhino, Archicad, and Vectorworks), interpreting prompts in the context of the model’s actual 3D geometry.

This matters most when geometry is still moving. During DD, an exterior elevation might shift three times in a morning. An integrated tool re-renders the new state in seconds. An export-based tool restarts the round trip every time.

Data residency is the other half of the equation. Keeping model data inside your CAD environment, with cloud processing handled through a vendor under enterprise terms, is a different posture than uploading screenshots to a public generative service. For firms with NDA-heavy project rosters, that posture matters.

Three habits for BIM managers adopting Veras

With over 60% of architecture firms now using AI in some form, the question has shifted from whether to adopt these tools to how to integrate them without disrupting BIM precision.

If you are a BIM manager folding Veras into a studio workflow, anchor it to three habits:

• Standardize prompt blocks at the project level the same way you standardize view templates. Lighting, lens, materials, mood, in that order.
• Use reference images before client meetings. Presentation decks need visual consistency, and using reference images is how you maintain it across multiple views.
• Reserve the final marketing pass for a photorealistic rendering engine. Veras is built for concept-to-DD speed, while V-Ray 7 and Corona 14 remain the tools for pixel-controlled final imagery. Enscape handles the real-time rendering layer between the two.

The pipeline that works in 2026: ideate in Veras, refine in Enscape, deliver in V-Ray or Corona. One ecosystem, no export-upload tax between phases.

→ Read next: Best AI rendering tools for architects 2026: 6 options compared

Integrated vs external AI rendering for Revit and SketchUp

Frequently asked questions

Do I need to export my Revit or SketchUp model to use Veras?

No. Veras runs as a native add-in inside Revit and SketchUp and reads your active 3D view directly.

Which Revit and SketchUp versions does Veras support?

Veras is currently compatible with Revit 2022, 2023, 2024, 2025, and 2026. Veras is compatible with SketchUp 2021, 2022, 2023, 2024, 2025, and 2026.

How does geometry override work in Veras?

This depends on the render engine you use. With render engine Version 7, geometry override is controlled in your prompt. Phrases like "strictly preserve massing" hold the AI to your model, while "reimagine the roof" gives room for creativity. With render engine Version 6 or below, use the Geometry Override slider in the Compose tab to handle this instead.

What is the render seed and when should I use it?

The render seed (available on engines V4–V6) locks the AI noise pattern so future renders inherit the same composition and atmosphere. It isn't available for render engine V7 onward, so you can use Reference Images instead. Simply feed a previous render back in as a reference to stay consistent while changing one prompt variable, or for producing a coherent set across a client deck.

Can Veras render only part of an image instead of the whole view?

Yes. Smart Selection lets you target a region, object, or material in an existing render and re-render just that area. Use Modify to change materials or finishes while preserving geometry, or Replace to discard the selection and reimagine it entirely. The rest of the image stays pixel-identical.

Is Veras a replacement for V-Ray, Corona, or Enscape?

No. Veras handles concept and design-development imagery using AI. V-Ray and Corona remain the tools for final photorealistic deliverables, while Enscape covers real-time rendering between the two phases.

Does Veras send my BIM model to the cloud?

Veras processes AI rendering in the cloud and requires an active internet connection. It reads geometry, camera, and material data from your active view, not the entire BIM model file. You can review Chaos’s AI policy here.