AI rendering does not begin with a prompt. It begins with what the model sees. The base image carries composition, proportions, camera, and the basic spatial relationships of the project.
Input types
- Sketch: useful for early mood, less useful for precision.
- Clay render: strong for controlled architectural visualization.
- 3D model screenshot: fast for internal testing.
- Existing render: good for refinement and atmosphere.
- Text only: useful for ideation, lowest control.
Best input by project stage
Concept work can start with a sketch. Client-facing presentation should usually start from a more precise 3D input, clay render, or existing visual.
Main claim: A better base image gives AI less room to misunderstand the architecture.