Architectural Design Brainstormer
Stuck on a design? Describe the site and program, get visual concepts to react to
Give it your project brief and constraints. It writes Midjourney/DALL-E prompts grounded in architectural precedents and spatial logic — not generic "beautiful building" prompts. Get 10 directions in the time it takes to sketch one.
The Problem
You have a brief, a site, and a deadline. You need concept options — not one, but several. The usual approach is browsing ArchDaily and Pinterest for a couple of hours, then sketching 2-3 directions. Most projects never explore more than that because there isn't time. Clients want to see options. Competitions demand range. But generating architectural concepts that respond to real constraints — program, site, climate, structure — takes effort that mood boards can't shortcut.
How It Works
Tell it: "200m² community library, Nordic climate, heavy timber, double-height reading room." It writes image prompts that specify spatial relationships, materials, lighting, and structure — not just "modern library building." Each prompt is tuned for Midjourney, DALL-E, or Stable Diffusion. It also suggests precedent projects worth looking at. Run it 5 times with different constraints and you have 20+ directions to react to before you touch a sketchbook.
"200m2 community library, heavy timber, Nordic. give me 5 directions""biophilic office lobby with double-height atrium and living walls"200m² community library, heavy timber structure, Nordic climate. Give me 5 directions.
Direction 1 — Compression Forest CLT columns at 1.5m grid create a dense interior grove. Reading alcoves carved between columns. Exposed glulam ceiling with acoustic felt hanging panels. Reference: Kengo Kuma's timber grids. Direction 2 — Stack and Cantilever Three offset CLT floor plates cantilevered 3m over the entry plaza. Each plate a different program level. North-facing clerestories for diffuse light. Snowload expressed as deep roof overhangs. Direction 3 — Buried Plinth Library sunk 1.5m into the earth, half-bermed. Green roof at grade level doubles as park. Timber above-grade volume reads as a pavilion within the landscape. Direction 4 — Lantern Heavy timber post-and-beam with translucent polycarbonate infill on south and west. Glows at dusk. Community beacon. Manages solar gain with operable timber louvers. Direction 5 — Stave Reference Vertical stave cladding (pine, charred on lower 1m) references Norwegian stave-church tradition. Compression-ring roof structure. Single open reading room, no internal walls above 1m.
Key Features
Who It's For
Architects in early design who want more options on the table before committing. Competition teams exploring multiple directions fast. Interior designers building concept presentations. Anyone stuck between the brief and the first sketch.
This tool generates text prompts for AI image generation — it does not produce images directly. Output quality depends on the AI image generation platform you use. Generated concepts are meant for early ideation and do not constitute buildable designs. Spatial relationships in AI-generated images may not be architecturally accurate. Always validate concepts through proper architectural development with licensed professionals.
Frequently Asked Questions
Do I need a Midjourney subscription to use this?
The Brainstormer generates text prompts that you can use with any AI image generation tool — Midjourney, DALL-E, Stable Diffusion, Adobe Firefly, or others. You need a separate account with your preferred image generation platform to create the actual visuals.
Can it help with client presentations?
Yes — the tool is specifically designed to generate presentation-quality concept visualizations. You can specify the presentation context (competition board, client meeting, design review) and the tool adjusts prompt complexity, framing, and mood accordingly.
How is this different from just using Midjourney directly?
The Brainstormer adds architectural intelligence to the prompt generation process. Instead of generic "beautiful building" prompts, it structures prompts around spatial programming, design theory, precedent analysis, and buildable architectural relationships — producing more architecturally meaningful results.
How do I generate better architecture concept images with AI?
The key is specificity. Effective architectural AI prompts include: spatial relationships ("double-height atrium with mezzanine level"), materials ("exposed board-formed concrete, weathered corten panels"), light quality ("north-facing diffuse light, deep shadow reveals"), scale reference ("human figures at ground level, 8-story building"), and a drawing type ("axonometric section perspective"). Generic prompts like "modern office building" produce generic results. The Brainstormer structures all of these elements from your brief automatically.
What are the best Midjourney settings for architectural visualization?
For architectural renders: use --ar 16:9 or --ar 3:2 for widescreen compositions, --stylize 300-750 for more interpreted artistic results, and add "architectural photography, Canon EOS, wide angle, professional" for photorealistic outputs. For concept sketches, add "hand-drawn architectural sketch, ink on paper, loose linework." For competition boards, use "--ar 1:1" or "--ar 2:3" for portrait boards and add "competition board layout, white background." The Brainstormer automatically applies the right parameters for your intent.
How do I use AI to explore multiple design concepts quickly for a competition?
Start by defining your design constraints: program, site orientation, climate, structure preference, and client values. Then generate 5-6 divergent concept directions by varying one major parameter each time — structure type, circulation logic, relationship to ground, material palette, or formal strategy. Use the Brainstormer to write a distinct prompt for each direction. Generate 3-4 image variants per concept, select the strongest, and refine with more specific prompts. This approach can generate 20+ concept directions in a morning, far more than traditional sketching allows.