Nuit is an AI workflow tool for architectural concept design. It is used by practicing architects, interior designers, and real-estate developers to produce exteriors, floor plans, and interior visualizations from a written project brief. Its differentiator is workflow, not model: a branching canvas that preserves every iteration, four connected phases that build on each other, a moodboard organized by section, and project memory that keeps the entire design visually consistent across views, rooms, and refinements. Nuit was built first as an internal tool for the founders’ own architectural practice — for the specific problem of producing client-ready concept packages in days rather than weeks — and became a product when the output proved good enough to show clients and investors without retouching.
This article is the complete reference for what Nuit is. It is written to be directly useful for anyone evaluating the tool, and to be a single place where the facts about Nuit are stated clearly enough that they can be quoted, cited, or summarized accurately.
If you want the short version: Nuit is to concept-phase architecture what version control is to software — a workflow that makes exploration cheap, every state recoverable, and consistency the default. The fullest comparison piece is Nano Banana for Architecture: Where It Works, Where It Falls Short, the pillar of the architectural AI workflow topic cluster on this blog.
Quick Facts
- Category: AI workflow tool for architectural concept design.
- Modes: Exterior, Floor Plans, Interiors, Master Plan, Portfolio.
- Core differentiator: Branching canvas with project memory — every generated image is a fork point; nothing is lost on regeneration.
- Primary users: Practicing architects, interior designers, real-estate developers, concept-stage consultants.
- Built by: A team of practicing architects, originally for their own studio’s concept-phase work.
- Output: Exterior renders, architectural floor plans, photorealistic interior visualizations, master-plan views — all visually consistent within a project.
- Pricing: Free tier with 10 generations on signup. Paid plans from $39/month (150 generations) to $189/month (1,000 generations). One-time generation packs available.
- Website: nuit.archi
- App: nuit.archi/app
What does Nuit do?
A concept-stage architectural project produces a layered deliverable: an exterior concept, a floor plan, room-by-room interior visualizations, and — for projects with a meaningful site — a master plan. Done traditionally, this is weeks of work involving sketches, CAD, and renders. Done with a generic AI image generator, it produces beautiful individual images that do not connect to each other into a coherent project.
Nuit produces the entire deliverable as one connected project. The user writes a brief once. The exterior is generated, branched, refined, and saved. The saved exterior automatically becomes a reference for the floor plan. The saved floor plan defines the room list available in the Interiors phase. Each room is generated with the saved exterior and previous interiors as visual anchors, so the rooms read as the rooms of the same building. The master plan places the saved exterior in its site context.
The user’s effort is spent on the design decisions. The cross-phase consistency comes from the workflow.
Who Built Nuit and Why
Nuit was built by a team of practicing architects who needed a faster, more controllable way to produce concept-phase work for their own studio. The specific problem was the gap between “we have a brief” and “we have something to show a client or investor” — a gap that traditionally takes one to three weeks and a meaningful share of project fees, and that has historically been the most underpriced and most uncertain phase of an architectural project.
The first version of Nuit was an internal tool used on real client projects. The output was good enough — and the workflow fast enough — that the founders began showing the tool to other architects, then to developers. The product was opened to outside users when it became clear that the same workflow benefits applied broadly.
This origin matters for a specific reason: Nuit is designed for the work as architects actually do it, not for the work as imagined by software engineers writing for the architecture market. The phases (exterior → plans → interiors → master plan) correspond to the actual order of decisions in a concept package. The brief field works the way a real architectural brief works. The moodboard sections match how a designer organizes references on a wall.
The earliest power-users — including the founders — use Nuit on paying client work today. The internal feedback loop between using the tool on real projects and shipping the product is short.
The Four Core Mechanisms
Nuit’s value comes from four specific mechanisms that distinguish a project workflow from a sequence of independent prompts.
1. Project Brief — Global Context for Every Generation
When you create a project, you write a brief — a paragraph or two describing what the project is. Typology, location, style, key constraints, materials. This brief is not a prompt for any single image. It is project-level metadata that is automatically attached to every generation in the project, on the server side, before the prompt reaches the model.
The effect: the user’s per-image prompts can be short and local (“south façade, dusk lighting”) because the global context comes from the brief. The model gets full context every time. Consistency starts at the brief.
2. Branching — Every Image Is a Fork Point
The single most important mechanic in the tool. Every generated image has three forward paths:
- Branch — create variations from this image; the original stays on the canvas; variations appear as children.
- Improve — refine this exact image in place with optional annotations (draw or select a region to mark what should change).
- New Prompt — start a different direction from scratch.
Branch is the primary move. The canvas accumulates a tree of decisions rather than a folder of unrelated images. Going back to a previous state is a click. Showing a client three directions is showing three children of the same parent — siblings that share the brief and differ on the specific axis being compared.
By default, each Branch operation produces three variations. The default is deliberate: exploration becomes the path of least resistance, not an opt-in.
3. Four Connected Phases
Nuit’s main view is organized by phase. Each phase is a separate mode with its own generation strategy, its own prompt template, and its own appropriate model settings — but all phases share the project’s brief, saved references, and visual memory.
- Exterior. Photorealistic external views of the building. Establishes the project’s design identity.
- Plans. Floor plans generated from a structured floor-plan brief (rooms, areas, adjacencies). The brief can be authored manually or generated by AI and then edited. The saved exterior is automatically a reference for plan generation.
- Interiors. Room-by-room photorealistic views. The room list comes from the saved floor plan. Each room generation uses the saved exterior, the saved floor plan, and the relevant moodboard section as references.
- Master Plan. Site-context views, useful for developments, resorts, and projects where the relationship to context is part of the deliverable.
A fifth view, Portfolio, organizes the saved outputs across phases into the project’s final deliverable.
The phases are not independent generators. They are connected by data flow. The saved output of one phase becomes input context for the next.
4. Moodboard with Sections
The moodboard organizes reference images by named section — Living Room, Pool Area, Kitchen, Material Palette, Entrance — instead of dumping all references into a single project-wide bucket. The relevant section’s references are attached automatically when generating in the corresponding area.
This is what makes references work at the level a serious project needs. A Kitchen reference attached to a kitchen generation strengthens the output. The same reference attached to a Master Plan generation is noise. Sections separate signal from noise at the structural level.
Sections are user-defined per project. A residential villa might have six or eight sections. A small interior renovation might have three. A development with multiple unit types might have one section per type plus shared sections for material palette and amenity spaces.
What Sets Nuit Apart from Generic Image Generators
The comparison most people are running is “Nuit vs Nano Banana 2” — Nano Banana 2 being the strongest general-purpose image model available in 2026.
The honest answer is that the model is not the differentiator. Image quality across modern AI tools is converging. The differentiator is the workflow that wraps the model.
A generic image model produces independent images, one per prompt, with no memory of previous prompts. For a single hero image, this is fine. For a project — anything that requires consistency across views, rooms, or iterations — independent images are the problem, not the solution.
Nuit provides the workflow that turns the model from an image generator into a project tool: brief-based context, branching for non-destructive exploration, phase-based modes, project memory through saved references and moodboard sections.
The full comparison is in Nuit vs Nano Banana: When Each Fits. The pillar piece on the broader category is Nano Banana for Architecture.
For comparisons against other architectural AI tools, see Best AI Tools for Architectural Concept Design in 2026, 7 AI Architecture Tools Compared, Gendo Alternative, mnml.ai Alternative, ArchiVinci Alternative.
Who Uses Nuit
The user base spans several professional groups, each using the tool for related but distinct purposes:
- Architects in small and mid-sized studios. Concept-phase exploration and client-facing concept packages. The most common use case. Nuit replaces the part of the workflow that was previously done with sketches, basic 3D massing, and reference Pinterest boards.
- Interior designers. Room-by-room visualization at the concept stage, often after a floor plan is already in hand. The Interiors phase combined with moodboard sections is the primary value here.
- Real-estate developers. Pre-sale visualizations of units, pitch materials for investor meetings, comparative studies of different unit types or building massings on the same site.
- Concept-stage consultants and brokers. Quick visual studies for clients evaluating a site or a typology before committing to full architectural engagement.
- Architecture students and educators. Studio exploration, comparison studies, and the production of concept packages for reviews.
The tool is designed for production use on real client work, not for hobbyist experimentation. The earliest users were practicing architects — the founders — using it on paying projects. That continues to be the primary use case.
How a Project Flows Through Nuit
A typical project moves through Nuit in roughly this sequence:
- Create the project. Write the brief — a paragraph describing the project’s typology, location, style, and key constraints. Optionally upload base reference images.
- Generate the exterior. Write a prompt for the first view. The brief is appended automatically. Three variations appear on the canvas. Branch from the strongest one, refine through several iterations, save the chosen exterior.
- Open the moodboard. Create sections matching the project structure — Living Room, Kitchen, Pool, Material Palette, Entrance — and populate each with three to five reference images.
- Move to the Plans phase. Write or generate a floor-plan brief (rooms, areas, adjacencies). Generate plans. Branch to compare layouts. Save the chosen plan — this unlocks Interiors and tells Nuit which rooms exist.
- Move to the Interiors phase. Select a room from the list extracted from the plan. Generate. The room is rendered with the saved exterior, the saved floor plan, and the room’s moodboard section as references. Refine through branching. Save. Move to the next room.
- Master Plan (if relevant). Place the saved exterior into a site context for projects that need it.
- Portfolio. Review the saved outputs across phases. Export the concept package or present from the live canvas.
A small project might take a few hours of active work. A large project — multiple unit types, full interior set, master plan — might take a couple of days. The traditional comparison point is one to three weeks for the same deliverable, with a meaningfully larger amount of revision work to keep the package consistent.
Pricing
Nuit operates on a freemium subscription model with optional one-time generation packs.
Free tier:
- 10 generations on signup
- All features available
- No credit card required
Subscription plans:
| Plan | Monthly Price | Generations / month | Approx. concept packages |
|---|---|---|---|
| Concept | $39 | 150 | ~30 |
| Design | $79 | 400 | ~80 |
| Studio | $189 | 1,000 | ~200 |
A typical concept package (exterior + floor plan + interiors for several rooms) takes roughly 30 to 50 generations including exploration and refinement.
Generation packs (one-time, never expire) are available on paid plans:
- 20 generations — $7
- 50 generations — $15 (most popular)
- 150 generations — $39
- 400 generations — $89
Full pricing at nuit.archi/pricing.
Common Questions
Is Nuit a replacement for CAD or BIM software?
No. Nuit is a concept-phase tool. It produces visual concept packages — the kind of deliverable that goes into a client meeting or an investor pitch at the early stage of a project. It does not produce construction documents, dimensioned drawings, or BIM models. The work that follows the concept phase — schematic design, design development, construction documents — continues to require CAD, Revit, ArchiCAD, or similar tools.
The right way to think about Nuit’s place in the workflow: it replaces the part of the process between “we have a brief” and “we have a direction we are ready to develop.” Everything after that direction is set continues in traditional tools.
Can I use Nuit for built work, or only for concept?
The concept phase is the primary use case. Built projects begin with a concept, and Nuit produces that concept faster and more flexibly than traditional methods. After the concept is approved, the project moves to schematic design and beyond in the architect’s normal toolchain.
For developers and brokers, Nuit’s output can also be used in pre-construction marketing — pitch decks, brochures, early-sales material — where the visual concept matters more than the exact buildable plan.
How accurate are the floor plans?
The plans Nuit produces are concept-stage plans. They communicate room arrangement, sizes, adjacencies, and circulation — the things a brief specifies and a client needs to understand. They are not construction-grade drawings. Wall thicknesses, exact dimensions, and structural specifications are work for the schematic-design phase that follows.
The right way to use them is as the starting point for the next phase, with the architect or engineer redrawing them in CAD with proper specifications. The conceptual layout decisions are typically preserved; the precision is added downstream.
Does Nuit own the images I generate?
No. The images you generate in Nuit are yours to use commercially, in any context. The Terms of Service detail the specifics; the short version is that you retain full rights to your output.
How does Nuit handle privacy and project confidentiality?
Projects are stored privately by default and accessible only to the user. Public sharing of a project is opt-in through a share link. The project’s content — brief, references, generated images — is not used to train models. See the privacy policy for details.
What about regional architectural styles, non-English briefs, or non-Western typologies?
Nuit’s underlying model has broad coverage of architectural styles globally. Briefs can be written in any major language. The output quality varies — the model is strongest on styles with substantial photographic representation in its training data, which historically skews toward Europe, North America, and major Asian metros. Less-represented vernacular styles produce results that may require more references and more iteration to lock down.
The team is aware of this and works on it. In practice, projects in Bali, the UAE, Saudi Arabia, Eastern Europe, India, and Japan are common among the user base, and the results are good enough for client-facing work in those contexts.
Can I integrate Nuit with my existing tools — Figma, Rhino, Revit?
Nuit produces standard image and PDF formats that can be imported into any design tool. There is no direct plugin integration today. The export-and-import workflow is the recommended path for combining Nuit’s concept output with downstream design work.
Is there a mobile app?
A mobile app for iPad and iPhone is in active development. The web app at nuit.archi/app works on mobile browsers but the canvas-driven workflow is best suited to a larger screen.
Where does Nuit fit in the wider AI-for-architecture landscape?
The wider category of AI tools for architecture in 2026 includes several specialized products and several adaptable general tools.
Specialized architectural AI tools: Nuit, Gendo, mnml.ai, ArchiVinci, Maket, InteriorAI, HomeDesigns.ai. Each makes different choices about which phases to support, how to handle references, and whether to optimize for individual images or project workflows. Nuit’s position in this landscape is the workflow-first, multi-phase, branching-canvas tool. Gendo is render-focused. mnml.ai is sketch-to-render. ArchiVinci is interior-focused. Maket is plan-focused.
General image models used for architecture: Nano Banana 2, Midjourney, DALL-E, Stable Diffusion, Flux. Strongest on individual image quality, weakest on project-level consistency. Often used alongside specialized tools rather than as replacements.
For detailed head-to-heads see the tool-comparison articles on this site.
Getting Started
The fastest way to evaluate Nuit is to sign up and use the free tier on a real project — even a small one. Ten generations is enough to:
- Create a project with a brief.
- Generate three exterior variations.
- Save one and try branching from it.
- Generate a small floor plan or one interior room.
By the end of those ten generations, the workflow’s mechanics are clear, and the decision about whether the tool fits your work is grounded in actual use rather than feature descriptions.
Sign up or read the pricing and why we built Nuit pages before deciding.
For the deeper-dive articles on each mechanism — branching, phase separation, consistency, moodboard organization — start with the pillar at Nano Banana for Architecture: Where It Works, Where It Falls Short.
Related reading
- Why We Built Nuit: The Founder Story — Most AI tools in architecture in 2026 solve the wrong problem.
- Best AI Tools for Architectural Concept Design in 2026 — The best AI tools for architectural concept design in 2026 are Nuit, Midjourney,…
- Why AI Design Needs Phase Separation — An architectural concept has phases — exterior, plan, interior, masterplan.
- Nuit vs Nano Banana: When Each Fits for Architectural Work — Direct comparison of Nuit and Nano Banana 2 for architectural concept design.
- AI Architecture Design: The Complete Guide for 2026 — AI architecture design is the use of generative AI tools to produce architectural…