5 AI Render Prompts Architects Use for Client Work
The prompts below aren't for moodboards or Pinterest. They're the structures we see working for professional studios use to produce concept renders, planning visuals, and client presentation imagery in Fenestra - tested across Flux 2 Pro, Nano Banana Pro, and Flux Krea Dev.
Shaun McCallum
February 9, 2026If you're producing AI renders for clients - RIBA Stage 2 concept work, schematic design presentations, planning submissions, or competition boards - vague prompts cost you billable hours. "Modern living room, photorealistic" produces a thousand interchangeable grey boxes, none of which you can put in front of a client.
A studio-grade exterior commission still costs £1,500–£3,000 in the UK and $2,000-$4000 in the US. The point of running this in Fenestra isn't to replace that work - it's to compress concept iteration from days to minutes so you can show a client three directions before lunch, then commission only the one they pick.
The fix is simple. Good prompts do three things:
1. Set the camera — where are we standing? What lens?
2. Name materials — not "modern" but "light oak flooring with brass fixtures"
3. Describe the light — time of day and mood matter more than furniture
Here are the 5 prompts I keep coming back to. Each one is copy-paste ready. If you want to follow along, [open Fenestra](https://fenestra.app/spaces) - you can swap between Flux, Nano Banana, and other models in one workspace to see which handles each prompt best.
I'd recommend using Text to Render and Flux Krea Dev for this but you can combine with other workflows.
Run this prompt in Fenestra
Create Now1. The Warm Living Room
The living room is the most requested interior render - and the one where "modern" as a prompt fails hardest. You need to pick a lane.
"Photorealistic interior photograph of a Scandinavian living room with double-height ceilings and floor-to-ceiling windows overlooking a pine forest. Light oak flooring, a cream boucle sofa with linen cushions, a round walnut coffee table, and a single dried eucalyptus stem in a ceramic vase. Late afternoon golden hour light casting long soft shadows across the floor. Shot on a wide-angle lens at eye level. Zoomed out showing all the details in the living room."
The camera position ("eye level"), the light source ("late afternoon golden hour"), and the material palette ("light oak", "cream boucle", "walnut") give the model enough to produce something that looks designed, not generated.
Try swapping the style too - replace "Scandinavian" with "Mediterranean" or "Industrial Loft" and keep the rest of the structure. The specificity of the materials and lighting carries across styles.
Start Designing Interiors
Try Yourself2. The Moody Bedroom
Bedrooms are all atmospheric. The trick is to tell the model what the light is doing - not just what furniture is in the room.
"Photorealistic interior of a boutique hotel bedroom at night. A king-size bed with rumpled white Belgian linen sheets, a cognac leather headboard, and brass reading lamps on either side. Floor-to-ceiling sheer curtains with a city skyline visible through the glass. Soft warm ambient light from the lamps — no overhead lighting. A single espresso cup on the oak nightstand."
"No overhead lighting" is a powerful negative instruction - it forces the model to create atmosphere from practical light sources, which always looks more realistic. And small details like "rumpled sheets" and "single espresso cup" stop AI from producing that sterile hotel-catalogue look.
This prompt works brilliantly with Nano Banana Pro if you're working from an existing image, or with Flux for text-to-render from scratch.
3. The Editorial Kitchen
Kitchens are technically complex - lots of materials, fixtures, and geometry. Being specific about handles, tiles, and countertops makes or breaks the result.
"Interior photograph of a luxury kitchen with floor-to-ceiling dark green marble slab backsplash, matte black cabinetry with integrated handles, and a waterfall-edge marble island. Brass pendant lights hang above the island. The countertop has a single cutting board with fresh lemons. Moody, dramatic lighting — like an architectural magazine editorial."
"Like an architectural magazine editorial" is a style cue that consistently produces higher-quality, well-composed outputs across every AI model I've tested. It shifts the render from "product shot" to "editorial" - better composition, better colour grading, better mood.
If you want something lighter, swap the palette: "white shaker cabinets, butcher block island, subway tile backsplash, morning light flooding through the window above the sink." Same structure, completely different feel.
Design Your Dream Kitchen
Try Yourself4. The House Exterior
Interior prompts get all the attention, but exterior renders are just as powerful for early design concepts and client presentations. The principles are the same - camera, materials, light.
"Photorealistic exterior photograph of a contemporary two-storey home at golden hour. Clean white rendered walls with dark timber cladding accents, a flat roof with hidden guttering, and a cantilevered upper floor. A mature olive tree in the front garden, concrete pathway leading to a pivoting front door. Long shadows on the lawn. Shot from the street at eye level."
Exterior prompts benefit from "shot from the street at eye level" - it's a natural perspective that avoids the drone-shot look most AI defaults to. And always include landscaping. A building without context looks like a 3D model, not architecture.
Want a totally different mood? Try this variation:
"Exterior of an English countryside cottage in autumn. Stone walls with climbing ivy turning red and gold. A slate roof, timber-framed windows with warm light glowing inside. Overcast sky with soft, even light."
Explore Facade Materials
Get Started5. The "Lived-In" Room
This is the prompt technique that made the biggest difference to my renders. Instead of describing more furniture, describe the feeling of the space.
"Wide-angle interior shot of an open-plan loft apartment. A modular sofa in charcoal wool, a six-seater dining table in light ash, and a kitchen island visible in the background. Polished concrete floors, steel-framed windows. Late afternoon light creates warm patches on the floor. The space feels lived-in — a stack of architecture books on the coffee table, a wool throw draped over the armrest, a half-drunk coffee left on the counter."
"The space feels lived-in" followed by specific details is the single most underrated prompting technique for interiors. It stops AI from producing showroom renders and starts producing spaces that feel real. A book left open, shoes by the door, a jacket draped over a chair - these small cues completely change the output.
This works with any AI model, but I've found Nano Banana Pro and Flux 2 Pro work well because of their conversational editing - you can start with a clean render and then add these details in follow-up prompts.
Every prompt above is structured for client-grade output. Open Fenestra, upload a sketch, plan, or massing model, and run any of these across Flux 2 Pro, Nano Banana Pro, and Flux Krea Dev side-by-side in one workspace.
Used by architecture and interior design practices across the US, UK, and Australia.
See pricing for studios
Start RenderingRelated Articles
Fenestra 101: Sketch to Render, Virtual Staging and AI Editing for Architects & Interior Designers
In this introductory Fenestra webinar, we walk through the core tools architects and designers can use to turn early ideas into visual concepts. The session covers sketch to render workflows, virtual staging, adding furniture, editing images, removing elements and using AI to quickly explore design directions.
Fenestra 101: Our First Webinar
We're running our first ever live webinar on June 3rd. A hands-on intro to Fenestra for architects and interior designers, covering image generation from prompts, 2D and 3D inputs, video, and where AI actually fits in your practice. 30 seats only.
Is conversation the next 3D modelling tool? We tested AI modelling in Rhino
We tested AI 3D modelling in Rhino, driving geometry by conversation, then rendering in Fenestra. Where text-to-3D works, and where it doesn't yet.
