The Photographer's Brief to AI Prompts for Architecture

The Photographer's Brief to AI Prompts for Architecture - This video created using Google LM Notebook illustrates how to use AI to generate accurate architectural renderings Generative AI image models cannot produce 100% dimensionally exact CAD drawings or measured geometry from text prompts alone, as AI image models approximate visual geometry rather than measuring it. However, to replicate a specific elevation as closely and consistently as AI tools allow, your sources outline a 2-part strategy using image-input workflows and strict elevation prompt formulas. 1. The Highest-Fidelity Method: Sketch / Image Input If you have a hand sketch, CAD elevation drawing, or 3D massing screenshot, uploading the image alongside a text prompt (or using ControlNet / BIM integration) is the single most accurate way to lock in the geometry. Template Prompt (with your drawing/sketch uploaded): "Turn this architectural elevation drawing into a photorealistic photograph of the finished building. Keep the exact same massing, floor heights, window grid, and proportions from the drawing. Materials: [Primary wall finish, e.g., off-white textured paint] with [Accent material, e.g., dark HPL cladding on staircase tower]. Lighting: flat overcast daylight for true material evaluation. Camera: straight-on elevation view, 35mm lens, verticals kept parallel, no keystoning. Photorealistic architectural photography." 2. The Text-Only Elevation Prompt Formula When generating elevations purely from text descriptions, follow a structured 7-layer photographer's brief: Formula: [Shot Type] + [Building Typology & Plot Dimensions] + [Architectural Style] + [2–3 Named Materials & Finishes] + [Lighting/Sky Control] + [Camera Controls] + [Realism Anchor] Example Copy-Paste Elevation Prompt: "Straight-on elevation photograph of a contemporary G+2 residential building on a 30 ft wide plot, 34 ft total height, flat roof with concealed parapet. Primary wall finish: off-white textured paint with dark grey ACP cladding on the staircase tower. Windows: powder-coated aluminium in charcoal. Lighting: flat overcast daylight to show true material colors and textures without harsh shadows. Camera: eye-level, 35mm lens, verticals kept parallel, no keystoning. Photorealistic architectural photography, clean facade, no people, no cars." 3. Key Rules to Maximise Elevation Consistency • Lock the Camera View: Specify "straight-on elevation view" and "verticals kept parallel, no keystoning". This prevents wide-angle distortion, angled perspectives, or converging vertical lines. • Control the Sky & Light: Use "flat overcast daylight". Direct or golden hour sunlight creates strong shadows that obscure facade geometry and distort true material colors, whereas overcast light gives an accurate evaluation of surfaces. • Name 2–3 Specific Materials: Use exact product and finish terms (e.g., "board-formed concrete", "charred timber cladding", "powder-coated aluminium mullions") rather than vague descriptors like "modern materials". • Lock Your Base Prompt: Keep fixed structural variables (plot frontage, height, window positions) identical in your prompt and only swap 1–2 material parameters for variations. • Iterate Conversationally: Use follow-up prompts (e.g., "same building, same camera angle, change only the cladding material to...") so the AI retains the context of the previous frame.