Claude for Architects: Write Client Proposal Visualization Prompts

You are an expert architectural visualization prompt specialist with 11 years of experience writing AI image generation prompts for residential architecture practices where client proposal visualizations must communicate spatial quality, material specification, and daylighting intent with enough accuracy that clients can make informed design decisions without misunderstanding what the completed building will look like. Help me create a scene composition prompt so I can build a repeatable generation workflow and produce a set of architectural visualization images for a residential project proposal that communicate the design intent accurately enough to support client sign-off without requiring a full 3D rendering production cycle. My situation: - Project type and proposal stage: [e.g., "a single-family residence in a coastal site — the proposal is at schematic design stage, presenting three design directions to the client before design development begins — the visualizations need to communicate spatial character and material palette rather than precise construction detail"] - Three design directions: [e.g., "Direction A: a low-profile rammed earth and timber structure with a single-pitch roof and deep overhangs — Direction B: a two-volume white render and black steel composition with a courtyard between volumes — Direction C: a board-formed concrete and glazing pavilion with a flat roof and a full-width terrace facing the ocean"] - Key spaces to visualize per direction: [e.g., "two views per direction — an exterior approach view showing the building's relationship to the site and landscape, and an interior living space view showing daylighting quality, material palette, and spatial proportion — six images total"] - Daylighting requirement: [e.g., "the site is south-facing with an ocean view to the west — the interior views should show late afternoon light entering from the west, creating warm horizontal light shafts across the interior surfaces and casting long shadows that reveal the spatial geometry"] - Material accuracy requirement: [e.g., "rammed earth must read as rammed earth, not as poured concrete — board-formed concrete must show the board pattern — timber must read as natural grain, not as painted or stained — these material distinctions are critical for client understanding of the design direction"] - Image tool: [e.g., "Midjourney v6 — the practice has an active subscription and the architect running the visualizations has intermediate Midjourney experience — the prompts need to be structured clearly enough that the architect can adapt them for minor design changes without needing to rebuild from scratch"] - Presentation format: [e.g., "images will be presented at A3 landscape in a printed proposal document and in a 16:9 Keynote presentation — the composition must work in both portrait-cropped and landscape formats without losing the key spatial elements"] Deliver: 1. Three exterior approach view prompts — one per design direction — each specifying the building massing description, the primary and secondary material surfaces, the landscape and site relationship, the time of day and sun angle for the approach view, and the viewpoint distance and height that communicates the building's scale relative to the site without distorting the architectural proportions 2. Three interior living space view prompts — one per design direction — each specifying the spatial dimensions in descriptive terms (long axis, ceiling height, depth-to-width ratio), the material surfaces visible from the viewpoint, the west-facing late-afternoon light quality, the shadow angle and length across interior surfaces, and the furnishing level (minimal furniture suggestion rather than fully styled interior) 3. A material accuracy specification block for each of the three material palettes — a Midjourney-readable description of rammed earth texture and color variation, board-formed concrete surface pattern and grey tone, and natural timber grain behavior under direct and diffused light — written with enough specificity that the model renders the material as distinct from visually similar alternatives 4. A daylighting language block — a reusable set of descriptors for the late-afternoon west-facing light condition, covering the light color temperature (warm amber, 2700K-3000K equivalent), the shadow length relative to object height, the contrast ratio between sunlit and shadow surfaces, and the sky color and cloud condition appropriate to a coastal site in late afternoon 5. A viewpoint consistency guide — for each of the six views, a specification of the camera height, the horizontal field of view, the distance from the primary building facade, and the foreground-to-background depth ratio — ensuring that the exterior and interior views for each design direction feel compositionally related rather than visually disconnected 6. A cross-direction visual differentiation review — a checklist for confirming that the three design directions read as clearly distinct from each other in the generated images, covering material palette contrast, roofline silhouette distinctness, interior spatial character difference, and color temperature variation — preventing the visual homogenization that occurs when AI visualization prompts are too similar across design alternatives 7. A prompt adaptation guide for schematic design changes — a structured process for updating any of the six prompts when the design changes during client feedback, covering which prompt elements to change for a material substitution, which to change for a massing adjustment, and which to preserve unchanged regardless of design development — so the architect can maintain prompt consistency across the full schematic design phase without rebuilding from scratch after every client meeting **Write every prompt as an architectural brief rather than an artistic direction — specify material, light, space, and viewpoint with the precision of a construction specification, and treat every visual ambiguity in the prompt as a risk that the model will resolve in a way that misrepresents the design intent to the client.**