Smart Prompt

AI Workflow / 2026
Prompt Builder — Vision Analysis · ComfyUI
Role — Concept, Development & Visualization

A render is only as good as its prompt, but writing one is slow and inconsistent—long strings of materials, lighting, mood, and camera terms, retyped from scratch every time and easy to get wrong. Smart Prompt removes that friction by reading the images instead of asking you to describe them.

You upload a design image and, optionally, a reference for mood. A vision model analyzes both—pulling materials from the design and atmosphere from the reference—and returns a set of editable, color-coded tags grouped by category. Those tags assemble into a clean, structured prompt in real time, which feeds straight into the render pipeline. The result is a prompt that is faster to build, easy to refine, and traceable back to the images that shaped it.

01
Upload Drop in a design image (required) and an optional reference. The design defines the materials; the reference sets the mood.
02
Analyze A vision model reads both images and pre-selects around twenty tags—materials, light & time, atmosphere, rendering style, camera.
03
Refine Tags are color-coded by source—blue from the design, green from the reference. Toggle, remove, or add your own to steer the result.
04
Generate The selected tags assemble into a structured prompt that runs straight through the render pipeline.