How Rendair AI handles scale and proportions

Answer

Short answer:

Rendair AI relies on your uploaded base images, such as 3D model screenshots, sketches, or elevations, to define and lock the scale of the generated output. While pure text prompts can be unpredictable regarding dimensions, using a visual reference ensures the AI adheres to the specific geometry, perspective, and proportions of your original design.

Overview

In architectural visualization, scale is not a suggestion; it is the defining constraint. Pure generative AI often struggles with this, frequently hallucinating incorrect door heights or distorting building massing to fit a square canvas.

Rendair AI solves this by separating "structure" from "style." When you upload a base image, the platform treats that image as the ground truth for geometry. The AI is directed to apply materials, lighting, and atmosphere onto the forms you provide, rather than inventing new shapes. This allows architects to maintain the precise proportions of a Revit or SketchUp model while benefiting from the speed of AI rendering.

How it works

The process depends on the input method you choose. The system prioritizes visual data over text descriptions when calculating scale.

For Image to Render and 3D Base to Render:

The platform analyzes the lines, contrast, and depth of your uploaded file. It maps the generated textures directly onto these recognized boundaries. If you upload a screenshot of a 3D model with a specific camera angle, the AI preserves that perspective. It does not "zoom out" or distort the aspect ratio; the output inherits the exact aspect ratio of your base image.

For Text to Render:

This mode generates images from scratch based on prompts. Here, scale is interpretive. The system uses standard architectural training data to estimate proportions (e.g., a standard door is roughly 2 meters high), but it is not metric-accurate. This mode offers specific aspect ratios (1:1, 2:3, 4:5, 16:9) and is best for mood boarding rather than documentation.

For Upscaling:

When increasing resolution (up to 8K), the system adds pixel density and refines details without altering the fundamental geometry. A 2-meter wall remains a 2-meter wall, just with sharper texture definition.

Capabilities

Rendair AI provides several tools specifically designed to maintain proportional integrity during the visualization process.

• Preserve aspect ratios: Any tool using a base image (rendering from sketch, 3D, or elevation) automatically outputs the exact same aspect ratio as the upload, ensuring no cropping occurs.

• Lock perspective: By using a 3D screenshot as a base, the vanishing points and horizon line remain exactly where you placed them in your modeling software.

• Orthographic interpretation: The system can render 2D elevations and floor plans without forcing them into a 3D perspective, preserving the flat, measurable scale of the drawing.

• Contextual editing: When using the editing brush to add elements (like a chair or tree), the AI analyzes the surrounding pixels to estimate the correct scale for the new object relative to the existing scene.

Inputs and outputs

To ensure the best handling of scale, the quality of the input is critical.

Inputs

• 3D Screenshots: JPG, PNG, or WEBP captures from tools like SketchUp, Rhino, or Revit.

• Drawings: Scans or exports of hand sketches and CAD elevations.

• PDFs: Single-page PDFs (converted to JPG is recommended for best results).

• File Size: Maximum upload size is 10MB.

Outputs

• Resolution: Default generations are roughly 1MP or 2K (depending on the model).

• Upscaling: Deliverables can be enhanced to 4K, 6K, or 8K for final presentations.

• Format: Images are generated in standard raster formats suitable for client presentations.

Limitations or notes

While Rendair AI is designed to respect the input structure, it is a generative tool, not a BIM engine.

• No metric awareness: The AI sees pixels, not measurements. It does not know a wall is 3000 mm long; it only knows it looks like a wall.

• Detail hallucination: At very small scales, the AI might misinterpret complex line work (like a dense railing) as a different material or object.

• Text-to-render variance: Generating images purely from text will rarely result in accurate architectural scale. This workflow should be reserved for early concept exploration.

• Precision services: For projects requiring 100% geometric accuracy guaranteed by human oversight, Rendair offers traditional 3D rendering services with defined revision rounds.