When should I use image-to-image instead of text-to-image?

Answer

Short answer:

Use image-to-image when you need to preserve the specific composition, structure, or perspective of an existing file, such as a 3D model screenshot, a hand sketch, or a site photograph. Text-to-image is better suited for early brainstorming when you do not have any visual assets and want to explore concepts from scratch.

Overview

In architectural visualization, the choice between these two modes depends on how much control you need over the physical geometry of the result.

Text-to-image relies entirely on your description. It is excellent for generating mood boards or abstract concepts where the specific layout of the room or building is flexible.

Image-to-image uses a visual reference as a "guide." It tells Rendair AI to keep the walls, windows, and furniture roughly where they are in your upload, but to apply new materials, lighting, and styles based on your prompt. This is the standard workflow for professionals moving from a massing model or sketch to a realistic visualization.

How it works

The process differs based on your starting point:

Text-to-Image

You provide a text prompt describing the scene. The AI generates a completely new image based on that description. You have control over the style and content, but the AI decides the camera angle and composition.

Image-to-Image

You upload a base file (a screenshot, sketch, or photo) and add a text prompt. Rendair AI analyzes the lines and shapes in your upload and "paints" over them. The result retains the aspect ratio and perspective of your original file while updating the visual aesthetic.

When to use image-to-image

For most architectural workflows, image-to-image is the primary tool once a project moves past the initial idea phase. You should use it in the following scenarios:

• You have a 3D model: If you use SketchUp, Revit, or Rhino, take a screenshot of your viewport. Image-to-image allows you to render that specific view without losing your design intent.

• You have a hand sketch: Upload a photo or scan of a drawing. The AI can interpret your lines as architectural elements and turn a rough sketch into a polished render.

• Renovation projects: Upload a photo of an existing room or building. You can prompt the AI to change the style (e.g., "modern scandinavian interior") while keeping the structural shell intact.

• Specific aspect ratios: Text-to-image in Rendair is limited to specific ratios (1:1, 2:3, 4:5, 16:9). Image-to-image inherits the exact aspect ratio of your uploaded file, giving you full control over the frame.

Capabilities

Using image-to-image allows you to maintain design consistency while automating the rendering process.

• Material application: Turn a white "clay mode" model into a textured building with brick, glass, and concrete.

• Lighting iteration: Take a daylight photo or render and process it to look like a night scene or a golden hour shot.

• Style transfer: Apply a specific artistic style or reference look to your existing geometry.

• Sketch interpretation: Convert loose line work into photorealistic images.

Inputs and outputs

Inputs

• Supported formats: JPEG, JPG, PNG, WEBP, TIFF, HEIC, AVIF.

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

• File size: Maximum upload size is 16MB.

Outputs

• Resolution: Default generations are approximately 1MP. Some modelsngenerate at 2K by default.

• Aspect Ratio: Matches the uploaded image exactly.

Limitations or notes

• Creative interpretation: While image-to-image respects your geometry, the AI may still reinterpret ambiguous shapes. A blurry sketch might be interpreted differently than intended.

• Detail accuracy: If your base image is very low resolution, the output may lack definition.