Evianis/travel-photo-abstraction

What it solves

This project provides a specialized skill for Codex to transform travel or everyday photographs into sparse, editorial-style abstract art. It avoids generic filters or decorative filler, instead focusing on preserving the structural essence of a photo—such as shapes, proportions, colors, and spatial rhythm—while reducing them to minimal abstract marks.

How it works

The skill analyzes a source photograph as a visual system, inventorying observable visual evidence before generating an abstraction. It maps these facts (like asymmetry, occlusion, and negative space) to minimal marks. To ensure accuracy, it uses a bundled reference library of 19 structural examples to guide the generation process. The final output is a a composition that combines the original photograph with the generated abstract panel, which is verified using Python scripts (via Pillow) for deterministic composition and validation.

Who it’s for

Users of Codex who want to create high-fidelity, structurally faithful abstract representations of their own photographs without relying on artistic guesswork or generic AI filters.

Highlights

• Structural Fidelity: Preserves spatial relationships, count groups, and focal coordinates rather than creating regular patterns.
• Evidence-Based Generation: Inventories visual facts before generating marks to avoid inventing content.
• Reference-Driven: Uses a specific structural reference index to select 2–4 relevant examples for each generation.
• Pixel-Verified Composition: Employs a Python-based finalizer to compose and verify the final artwork.