AI Visualizer

Making generative AI intuitive for architectural design

Bringing AI-powered visualization into the architectural workflow, helping designers explore ideas, control generated results, and iterate faster.

Graphisoft AI Visualizer integrates generative AI into Archicad, allowing architects to transform model views into visual concepts using text prompts and adjustable parameters. My work focused on making this emerging technology understandable and controllable through interface design, while contributing to the refinement of the generation experience. As the product evolved into AI Visualizer 2.0, the team introduced faster generation, reference-image guidance, and high-resolution upscaling to support a more fluid ideation process.

Year

2023

2023

Role

Product Designer

Product Designer

Work type

AI Product

AI Product

Project scope

Interaction Design, UI Design, AI Parameter Controls, Prompt Optimization, Design System Integration

Collaboration

Design, development, and cross-functional product teams

01 — The Challenge

Making AI generation understandable and controllable

Generative AI opens up new possibilities for architectural exploration, but its outputs can be difficult to predict. Architects need to understand how their input affects the generated image and how to guide the AI toward a useful result without losing the freedom to explore alternatives.

The challenge was to make the AI generation process feel more transparent and approachable within an established professional application.

Key design challenges

  • Unclear parameter behavior: Users needed to understand which settings influenced which aspects of the generated image.

  • Balancing creativity and control: The interface needed to support both experimental exploration and results that stayed closer to the original design intent.

  • Integrating into Archicad: The new AI experience had to fit the existing interface conventions and visual system.

  • Supporting iteration: The workflow needed to make it easy to experiment with prompts and settings, compare directions, and refine results.

02 — Understanding the Workflow

From an Archicad model to an AI-generated concept

The workflow starts with the architect's existing model, rather than a blank canvas.

  1. Prepare the model: The user creates a composition in Archicad using architectural elements or other modeled forms.

  2. Capture the view: The user opens AI Visualizer and captures the current model view.

  3. Describe the vision: A text prompt communicates the desired architectural style or atmosphere.

  4. Adjust the parameters: Controls let the user influence the AI's creativity and how closely it follows the original composition.

  5. Generate and explore: The system produces an architectural visualization that can inspire further design decisions.

My focus was on making the controls in this process understandable, so users could develop a more intuitive relationship between their inputs and the generated results.

03 — Design Approach

Making complex controls feel intuitive

As one of three designers on the project, I worked on the interface for AI generation parameters, including sliders and settings that influence the output.

Rather than treating the controls as isolated technical options, the goal was to make them understandable within the context of the creative task.

Clarifying parameter relationships

I designed controls to help communicate how different settings influence the generated image, making it easier for users to experiment intentionally rather than relying entirely on trial and error.

Balancing creative freedom and fidelity

The experience needed to support different levels of AI interpretation. Users could explore more imaginative results or aim for outputs that remained closer to their original composition. The interface had to make this balance accessible.

Working within an established design system

The AI Visualizer interface needed to feel like part of Archicad, not a separate application. I worked within its existing visual conventions and technical constraints, balancing usability improvements with consistency across the wider product.

04 — Improving the Generation Experience

Refining the output, not just the interface

The interface was only one part of the experience. The quality and relevance of the generated images also influenced how useful the tool felt to architects.

I collaborated with the team on refining generation prompts, including negative prompts, to help guide the AI toward better-quality architectural visualizations.

This required considering how the user's description, the underlying prompt structure, and the model's interpretation worked together.

The goal was to make the generated output more useful as a source of inspiration, while preserving room for unexpected ideas.

05 — AI Visualizer 2.0

From a capable AI tool to a faster, more flexible workflow

Generative AI technology continued to evolve, creating opportunities to improve the speed, flexibility, and quality of the experience. In AI Visualizer 2.0, the team moved from AUTOMATIC1111 to ComfyUI, a node-based workflow platform, while continuing to use Stable Diffusion for image generation.

The upgrade introduced three major capabilities.

Rapid Mode

Generation speed was a major barrier to iterative exploration. The new Rapid Mode dramatically reduced generation times, allowing users to test more ideas in less time.

The project presentation reported reductions from approximately 30–40 seconds to 1–2 seconds on faster machines, and from around six minutes to approximately ten seconds on slower machines.

Design value: Faster feedback supports a more fluid creative process, making experimentation less disruptive.

Reference Images

Users could guide the generation process with a reference image in addition to text prompts and parameters.

Instead of having to describe every visual characteristic in words, architects could provide a visual example of the style they wanted to explore.

Design value: Visual references provide another way to communicate intent and make style exploration more direct.

High-Resolution Upscaling

A new upscaling model could increase image resolution by up to four times within seconds.

Design value: Users could move from initial exploration toward more detailed, higher-resolution results without introducing a lengthy additional step.

Together, these improvements expanded the workflow from generating initial concepts to exploring visual styles more quickly and producing more detailed imagery.

My documented contribution was primarily focused on the original AI Visualizer's interface, parameter controls, and prompt refinement with the team. The 2.0 features are presented here as the evolution of the product; I would only attribute individual feature design to myself where my direct involvement can be confirmed.

06 — Design Constraints & Collaboration

Designing for a complex technical product

The project required close collaboration between design and technical teams. Interface decisions needed to account for how the generation system worked, while the underlying technology had its own limitations and performance characteristics.

The established Archicad design system also shaped what could be changed and how new controls could be introduced.

My contribution sat at the intersection of UX, UI, and technical collaboration: translating complex AI behavior into a more understandable interface while working within the constraints of a mature desktop application.

07 — Outcome

Bringing generative exploration into the architectural workflow

AI Visualizer brought AI-powered image generation directly into Archicad, giving architects a new way to explore visual directions from their existing models.

The original product focused on making generation controllable and accessible. AI Visualizer 2.0 expanded that experience with faster iteration, visual reference inputs, and high-resolution upscaling.

Together, these capabilities support a workflow in which architects can move between modeling, visual exploration, and refinement without leaving their primary design environment.

For me, the project explored an important product design challenge: how to make a powerful but unpredictable AI system feel understandable, useful, and responsive within a professional creative workflow.

← all work

Graphisoft · Archicad 27 · 2023

AI Visualizer

Generative AI for architects, back when it still needed a terminal.

What

Turns a simple mass model in Archicad into images of a building, for early visuals and ideas without detailed modeling.

My part

One of two designers on the interface. I did the UI ideation and prototyping, and we tested it together with 4 architects.

Result

Won the GRID hackathon (Business). Shipped 16 Nov 2023 as a free add-on for Archicad 27. Now a product line, up to 3.0 in Archicad 30.

Role

Product Designer, side project next to my main work

Team

2 designers on the UI, 1 designer on the AI pipeline, the dev team

Year

2023

Scope

UI ideationPrototypingUser testing

01 — background

Stable Diffusion was usable, but only for enthusiasts

Early in 2023, generative AI was suddenly everywhere, and we wanted Archicad to be part of it from the start. Getting there wasn't simple. I set up Stable Diffusion on my own machine through the Terminal and AUTOMATIC1111, the go-to interface back then: a single page packed with technical parameters. Each image took several minutes, and you never knew what you'd get.

We wanted architects to skip all of that. The AI would live inside Archicad: block out a few volumes in the model, and get back a picture of a building. No extra app, no command line.

What Stable Diffusion looked like in 2023, and what we kept out of Archicad.

What Stable Diffusion looked like in 2023, and what we kept out of Archicad.

02 — from many to one

From a Creative Day to a shipped add-on in six months

Creative Day is a voluntary day organised by our design team. Four of us used it to see how far AI rendering could go in Archicad: images from sketches, from site photos and from mass models, facade variations, interiors. For the GRID hackathon we kept just one direction: turning a mass model into inspiration. It was the easiest to build and the one people would get the most out of. Reworking an existing render made the AI invent details that weren't there, while a handful of simple boxes could turn into something rich: realistic to build, and worth having.

12/5

Creative Day

8 AI ideas tested in a day.

GRID

Hackathon

4 days. Figma prototype + working demo. Won Business.

4×

Design + testing

Narrowed the scope, tested with 4 architects.

16/11

Launch

Free experiment for Archicad 27.

↑ shipped!

AC28

Product line

Moved to the cloud, in paid plans.

03 — key decisions

Two or three clicks to an idea

The first concept wanted to cover everything: interiors, styles, weather, seasons, time of day. For version one we cut that back on purpose. The tool was meant to spark ideas, not to produce client-ready images, art styles or finely tuned results.

a

Cut it down to three controls

Prompt, shape fidelity and image size. The extras stayed on the list for later.

b

Name the control after the architect's question

“Creativity” describes the AI. Shape fidelity describes what the architect actually wants to know: how closely will the result follow my model?

creativityshape fidelity
c

Make an empty prompt still work

Every input is wrapped in a hidden prompt that already asks for a photo-quality building. An empty field still returns a building, and off-topic text can't pull the result off course.

d

Speak the architect's language

Model terms became plain words or disappeared behind good defaults. Developers got the exact model value behind every control.

StepsCFG ScaleModel ScannerIterationsPrompt strengthgood defaults
e

Errors that say what to do next

The engine ran locally, could fail to start, and was sometimes shared. The dialog explains what went wrong and warns before anyone stops a shared engine.

GRID concept · May 2023Shipped · Nov 2023

GRID concept, May 2023 → shipped dialog, November 2023.

04 — testing

The help was there. Almost nobody saw it.

To get good results, the model needed things architects rarely add, like ground under the building and a tree for scale. The prototype explained this three ways: an illustration, an example prompt and hover tips. We tested it with four architects. The results landed; the help mostly didn't.

3 / 4

missed the hover tips completely.

1 / 4

followed the illustration.

3 / 4

misread shape fidelity. One dragged it to 0%, expecting a more building-like image.

They fixed the model, not the prompt

When a result missed, they edited the geometry instead of the text. Shape fidelity turned out to be the control that mattered most.

“This is cooler than the last one!”

A tester, after adding ground in her third try

We didn't fix it with better onboarding. The developers changed how the model view reaches the AI, so ground and trees were no longer needed. The tutorial never shipped, and the step people kept missing was simply gone.

05 — the reception

Out in the world, for free

On 16 November 2023, AI Visualizer went out as a free experiment. Graphisoft's announcement put it simply: “you don't need to learn prompt engineering.” AEC Magazine, Architosh and Engineering.com all wrote about it. Feedback from the community was mixed, and refreshingly honest.

Installing it was the hardest part

The engine took 16–20 GB, system drives filled up, and Mac installs failed. Plenty of people never saw a single image. That was the price of running AI locally in 2023.

Not everyone saw the value

Some dismissed it as a “3D mood board” and felt the images strayed too far from their model.

Put to work by professionals

One firm ran it on a live project, comparing the convenience of staying in Archicad with more powerful standalone tools.

Used exactly as intended

A teacher watched weak student models become “excellent results”. One user even suggested calling it “AI Inspiration”, which was the whole point.

06 — where it went

From experiment to product line

In Archicad 28 it moved to the cloud: nothing to install, included in paid plans. Our local version was retired at the end of 2024, but parts of our dialog are still there: Lock result, Iterations (default still 35), Prompt strength and the 800×600 default size. A new version has shipped with every release since, up to 3.0 in Archicad 30.

Its headline feature, Image from composition, is still our original idea: the model sets the brief, the prompt sets the style.

note:

I worked on the experimental first version, as an optional project next to my main work. Once it became a product, the team responsible for that area took it further.

07 — what I learned

What I took from it

Users steer with what they know

We designed around the prompt. Architects steered by modeling.

Visible isn't the same as seen

The advice was correct, and almost nobody read it. Removing the need for it worked better than explaining it.

Designing for AI, not with it

In 2023, AI was the material we shaped into a product, not a tool that helped us design.

08 — beyond the product

Taking it off the screen

Later we took the same technology off the screen and made it interactive: instead of typing a prompt, visitors arranged blocks and objects in front of a camera, and the AI turned them into architecture live.

beyond the productAI Design Live Experience The same technology, off the screen and interactive.

next: your team?

Right now

Looking for a mid-senior product design role in B2B SaaS or fintech. Available from December 2026.

© 2026 Karola Menyhárt · Budapest

Built in Framer