AI Tools for the Concept Stage
The AI tools that help designers explore mood, massing and first ideas before they commit.
This guide groups AI tools for concept design by what they start from, which is a prompt, a sketch or model, or a real site. It uses each vendor's own documentation to say who each tool suits. You also get a kit to judge any tool on range, control and scale, and three steps from picture to plan.
AI tools for concept design are worth using for range. The more directions a designer sees before committing, the better the one they choose. This guide groups the main tools by what they start from, which is a prompt, a sketch or model, or a real site. It uses each vendor's own documentation. None of these tools makes a decision. The good ones make better questions for the first client meeting.
What the concept stage needs
The concept stage needs range, speed and atmosphere. Its trap is a picture with no scale. A client falls for an image that nobody can build at the size it suggests.
- Text-to-image. Input: A prompt. Output: Mood images. Typical risk: No scale, no plan.
- Sketch or model to render. Input: A sketch or 3D model plus a prompt. Output: Images that keep your geometry. Typical risk: Drift from the input.
- Site massing. Input: A site and simple volumes. Output: Massing with sun, wind and noise. Typical risk: Little atmosphere.
A kit to judge any concept tool
Give every tool the same brief, the same prompt text and the same hand sketch, with one plan and one perspective. Allow three generations each. Two designers should rate the results apart, then compare notes.
A sample brief. A 6 m x 9 m exhibition stand, open on two sides, for a fictional maker of acoustic panels. It has a 4 m height limit, a demo area, a meeting space for six and 4 m² of lockable storage. The materials are timber and felt, and the mood is calm and tactile.
What to record. Note the version, the plan tier and the price on the day. Time the first usable image and count the prompts it took. Note whether the tool accepts a sketch or model, and whether it holds the sketch's geometry. Check whether any output carries scale. List the export formats and the license terms for commercial use.
Rate each criterion from 1 to 5.
- Range: Distinct directions, not variations on one
- Control: Materials, light and composition change on request
- Faithfulness: The sketch's geometry survives
- Scale: Output carries real proportions or dimensions
- Handover: Output moves to a to-scale tool without redrawing from scratch
Text-to-image with Midjourney, Adobe Firefly and Krea
These tools are for mood boards and atmosphere. None accepts a plan, so none knows the stand is 6 m x 9 m.
Midjourney
Midjourney is the usual benchmark for light, material and mood. Its style references hold a look across a series. Its default model changes often, so note the version you use. Check its terms for commercial use on your plan.
Adobe Firefly
Adobe states that its Firefly models are trained on licensed content, which matters for client work. It also offers an architecture generator. Firefly now hosts partner models as well, so check the terms for the model you pick.
Krea
Krea puts several image models in one workspace. That includes real-time generation, which redraws the image as you change the prompt. It suits quick mood exploration.
Sketch and model to render with Veras, LookX, mnml.ai, Gendo and Rendair
These tools keep the designer's geometry. Judge them on how faithfully they hold the input sketch, and how much control they give over materials.
Veras
Veras renders from the live model as a plugin for Revit, SketchUp, Rhino, Archicad and Vectorworks. Chaos acquired its maker, EvolveLAB, in February 2025. It suits designers who already model in one of those programs.
LookX
LookX is an AI rendering platform for architects that works from sketches and models. It suits early design studies in the browser.
mnml.ai
mnml.ai turns sketches and model views into renders in the browser. It suits fast restyles of a view you already have.
Gendo
Gendo is an AI rendering and image-editing tool for design studios. It suits teams that refine one image many times.
Rendair
Rendair renders sketches and model views in the browser. It suits quick concept images from rough input.
Renderers inside your design software
Their advantage is that they start from your model, so scale survives.
SketchUp AI Render
It launched in December 2025 as part of SketchUp AI. It was formerly called SketchUp Diffusion.
Archicad AI Visualizer
Graphisoft's AI Visualizer renders from the Archicad model. Graphisoft switched version 1.0 off on 9 October 2026, so use the current version.
Vectorworks AI Visualizer
The AI Visualizer generates images from prompts inside Vectorworks.
D5 Render AI tools
D5 Render adds AI features to a real-time renderer that already holds your model.
Site and hall with Autodesk Forma and Monet
Autodesk Forma Site Design
Forma puts massing on a real site. It adds sun, wind and noise analysis that no image provides. It suits buildings, not stands inside a hall.
Monet
For stands, Monet generates a booth from a prompt, to scale, on a real venue plan. So the concept already sits in the hall. For pure mood, a text-to-image tool gives more range. Monet suits the step where the idea has to fit the hall.
From picture to plan
The handover is where concept work is wasted. Do three things before anyone falls for a picture.
- Freeze the chosen image. Save it with its prompt and version, and stop generating.
- Write down the decisions it implies. Note the heights, materials, openings and structure. A curved felt wall 3.5 m high is a decision, not an atmosphere.
- Redraw to scale. Move the decisions into a to-scale tool. Check them against the brief of 6 m x 9 m and 4 m high.
The next stage is covered in AI tools for schematic design.
Widen the field, then commit to scale
Use concept tools to widen the field. Then move the chosen idea into a to-scale tool before anyone falls in love with a picture.
Frequently asked questions
What is the best AI tool for concept design?
It depends on what you start from. For mood from text, try Midjourney, Adobe Firefly or Krea. To keep your own sketch or model, try Veras, LookX or a renderer inside your design software. For site and sun, use Autodesk Forma. For a stand that must fit a real hall, Monet generates a booth to scale.
Can AI concept images be used commercially?
Check each tool's terms for your plan. Adobe states that Firefly is trained on licensed content. Other tools set their own terms for commercial use, and those terms change. Record the version of the terms you read on the day you use the image.
Do AI concept renders show real scale?
Text-to-image tools do not, because they have no plan or units. Sketch-to-render tools and renderers inside design software start from your geometry, so proportions hold better. Massing tools work on a real site. Measure any image against the brief before it reaches a client.
To see your next stand concept to scale in a real hall, try Monet free and score it with the same sheet.