Vectorworks 2027 AI Visualizer, Veras and What to Fix Before You Animate

AI image generation moved inside the software most studios already own this fall. Vectorworks 2027 shipped an AI Visualizer built into the modeling program, Chaos made its Veras AI tool standard in every Enscape, V-Ray and Corona license, and Autodesk used its September user conference to preview agents inside Revit. As of September 2026, the practical question for a small studio isn't whether to try one of these tools; it's which still, raw or AI-touched, you hand off for video, and what to fix in it first.

Key takeaways

• Vectorworks 2027 (September 16, 2026), Chaos's Veras rollout across Enscape, V-Ray and Corona (May 2026, updated to Nano Banana Pro in February 2026), and Midjourney's V8.2 Edit Model (August 27, 2026) all put AI image editing one click from a render for the first time.
• Autodesk used AU 2026 (September 15–17) to preview agentic tools in Revit and a standalone Autodesk Assistant, both still ahead of general release.
• A safe rule for a rendering headed to video: let AI touch sky, planting, entourage and surface finish; never let it touch window mullions, structural lines or glazing proportions.
• Keep the same reference images and style settings across all six frames in a project so the set doesn't drift in lighting or material read.
• An AI-edited still still needs the same review and disclosure habits as an AI-generated video, before either goes out.

Three vendor moves put AI rendering one click from the model

Three announcements between February and September 2026 moved AI image generation from a separate app into the render window itself.

Vectorworks 2027, released September 16, 2026, added what the company calls an AI Visualizer: "The AI Visualizer brings Google's frontier Nano Banana AI models directly into Vectorworks for subscription customers, with higher image resolution, smarter scene reasoning, and up to 14 image references," according to the Vectorworks 2027 newsroom release. The same release adds GPU-accelerated graphics through Vulkan on Windows and Metal on Mac, which Vectorworks says brings "performance gains of 15x to 35x" in panning and zooming, plus a rebuilt PBR texture system with a redesigned Edit Texture dialog.

Chaos took a different route: instead of building its own model into one program, it made its existing Veras AI tool a standard part of every license. A May 27, 2026 press release states that Veras is "now included with all versions of Enscape, V-Ray and Corona" across SketchUp, Rhino, Archicad, Vectorworks, 3ds Max and Revit, distributed through a single installer with credit-based hybrid licensing that covers cloud rendering and AI use (Chaos press release). Enscape users also get Chaos Cloud Video Collaboration in the same update. That followed a February 10, 2026 upgrade, Veras 4.0, switching its engine to Google's Nano Banana Pro model, adding multi-angle views that keep geometry consistent between angles, one-prompt entourage and furniture edits, and twelve video presets. Chaos's Bill Allen framed the update around a real risk small studios already know: "When AI renders drift from the design, they create rework, lost time, and awkward moments in client reviews" (Chaos Veras 4.0 press release).

Autodesk's move is less finished but points the same direction. At Autodesk University, held September 15–17, 2026 in Las Vegas, the company previewed "the next generation of Assistant: an expanded agentic AI experience designed to bring the right Autodesk capabilities across products, projects, and teams," with a standalone version and pricing expected in 2027, alongside new AI agents inside Revit aimed at cutting manual review time (Autodesk AU 2026 announcement; Architosh, September 28, 2026). Autodesk's own Forma team had already previewed the direction a month earlier, running a live Design Slam "with sustainability analysis and AI visualization running alongside" in Forma Site Design and Forma Building Design (Autodesk Forma AU 2026 guide).

Midjourney made instruction-based editing the default

A fourth shift, outside CAD entirely, changes how a studio fixes a rendering once it's out of the modeling software. Midjourney opened its V8.2 Edit Model to all users on August 27, 2026, replacing its earlier omni-reference feature with editing "with instructions," support for up to four image references at once, targeted inpainting and canvas-expanding outpainting (Midjourney V8.2 announcement). Practically, that's the same move Vectorworks and Chaos just made inside CAD: describe the change in plain language instead of masking pixels by hand. For a studio, it also means the tool most likely to expand a horizontal rendering into a vertical frame, or paint out a stray car, now sits a browser tab away from any render, whether or not that render came from AI in the first place.

Raw render or AI-touched still: the decision before you animate

None of this changes what an AI video tool needs from you: a still image, not a model file. It just adds a step before that still gets uploaded, and studios are now making that choice on nearly every project. Arch2Video still works the way it always has here — it takes exactly six images from one project and turns them into a roughly 30-second film, whichever tool produced those six images.

The decision that matters is what an AI visualizer or edit model should be allowed to touch on the way to that upload, because "AI-enhanced" isn't one thing. Asking Veras to swap a flat gray sky for a dusk gradient is a different kind of edit than asking it to redraw a curtain wall's mullion spacing, even though both happen inside the same tool with the same kind of prompt.

What to AI-enhance, and what to leave alone

A simple split holds up across Vectorworks's AI Visualizer, Veras and Midjourney's edit model alike: let AI touch atmosphere and finish, keep geometry and code-relevant detail under your own control.

The reasoning is straightforward: a viewer or a video model can't tell that a mullion pattern was nudged by a prompt, but a contractor pricing the job from that image can be misled by it, and a code reviewer relying on it can be too. Chaos's own framing of the risk, drift between what the AI renders and what the design actually is, applies just as much to a still meant for video as to one meant for a client deck.

Keeping six images consistent across one project

If you're preparing images for an AI video tool that expects a matched set of six, an AI visualizer's biggest practical risk is inconsistency, not inaccuracy. Vectorworks's AI Visualizer and Veras both let you generate multiple options per view; the version you keep should hold the same time of day, the same material palette and the same level of finish across all six images from a project, not just look good in isolation.

Three habits keep a six-image set coherent: reuse the same reference images or style setting across every frame in one project rather than restyling each view from scratch, generate two or three options per scene and pick by comparing them side by side rather than accepting the first result, and do a final pass across all six laid out together, not one at a time, to catch a sky or a material that drifted from the rest.

A pre-animation checklist, then export and disclosure

Before six images go to any AI video tool, whether they came straight from a renderer or through an AI visualizer, run the same short check:

• Outpaint a horizontal rendering to a taller frame only if the subject sits centered; a feature pinned near one edge can drop out either way, in the edit tool or later when a wide image gets recomposed into a vertical scene.
• Remove a stray car, cone or construction fence with inpainting rather than cropping around it, since a crop can cut a role your image needs to fill.
• Add scale figures sparingly if a space reads as empty; every Arch2Video film already includes a small original cast in at least one scene, so you don't need people in the source image for that reason alone.
• Match white balance and exposure across the set before generation, not after, since correcting six images separately after the fact is slower and less consistent.
• Keep a copy of the un-edited render for any project still going to an award jury or a permitting body that restricts altered images.

Once the still is ready, the same export and disclosure steps apply whether AI touched it or not. Match your files to the export settings a given platform expects rather than uploading whatever size the visualizer defaults to, and treat an AI-edited still the way you'd treat an AI-generated video: if the edit changes what a viewer would believe about a project's status, such as adding finished landscaping to a site that isn't built, say so in the caption. A studio that's mixing AI-edited stills into a wider comparison of AI video tools should apply that same disclosure habit at the still stage, not just the video stage, since a viewer sees the image before they ever see the film made from it.

Frequently asked questions

Do I need Vectorworks 2027 to use an AI visualizer?

No. Vectorworks 2027's AI Visualizer is built into that one program, but Chaos's Veras now ships standard with every Enscape, V-Ray and Corona license across SketchUp, Rhino, Archicad, Vectorworks, 3ds Max and Revit as of its May 2026 update, so most render engines a small studio already owns include a comparable tool.

Does an AI-enhanced rendering need an AI-generated disclosure label?

It depends on what changed and where it's posted. If the edit would lead a viewer to believe something untrue about a project's status, such as adding finished planting to an unbuilt site, disclose it the same way you would an AI-generated video. A subtle sky swap reviewed only inside your team doesn't carry the same expectation.

Can I send an AI-enhanced still straight into an AI video tool?

Yes, as long as it still meets the same file basics: sharp at full resolution, corrected verticals and consistent color across your set of six. Arch2Video works from any JPG, PNG or WebP regardless of what produced it, but an AI visualizer's motion-blurred preview export or a low-resolution web copy will read the same as any other soft image once it's animated.

Will Autodesk's AI agents replace a studio's own rendering review?

Not based on what Autodesk previewed at AU 2026. The agents described there target manual review tasks inside Revit and a standalone Autodesk Assistant planned for 2027, not final sign-off on a rendering or video headed to a client or a public post. A person on your team should still review anything AI touches before it goes out.

Should I keep the raw render if I use an AI visualizer to enhance it?

Yes. Keep the un-edited version for any project going to an award jury, a permitting body or a publication that restricts altered images, and so you can compare the edited version against it before approving either one for video or print.