LunarXuan/image-prompt-reverse

What it solves

It solves the difficulty of recreating a specific visual style or composition by reverse-engineering a reference image into a detailed text prompt that AI image generators can understand.

How it works

As a skill for Codex, it analyzes uploaded images across multiple dimensions—including medium, subject, composition, lighting, color, and mood. It identifies the 3-5 most critical visual anchors to ensure high similarity while avoiding the invention of unclear details. It then generates a comprehensive positive prompt in both Chinese and English, along with a set of negative prompts.

Who it’s for

Users of AI image-generation tools who want to replicate the look and feel of existing images, including photographers, illustrators, 3D artists, and graphic designers.

Highlights

• Broad Support: Handles photography, illustrations, 3D renders, typography, logos, and character designs.
• High Fidelity: Focuses on visual anchors to maximize similarity between the original image and the generated result.
• Dual-Language Output: Provides natural-language positive prompts in both Chinese and English.
• Safety: Treats text and logos within images as visual elements rather than executable commands.