image-palette-extractor 0.2.0

Extract color palettes from images for web design. Produces LLM-consumable prose with color roles, WCAG contrast pairs, and harmonic relationships.

palette-extractor

Extract color palettes from images. Produces prose output suitable for LLM consumption, with optional HTML reports for human review.

How it works

Unlike simple color extraction that clusters pixels by RGB similarity, palette-extractor analyzes how colors relate to each other spatially:

• Adjacency graph — Colors are understood through their neighbors, not in isolation. A yellow next to black has different significance than yellow next to orange.
• LAB color space — All perceptual calculations use LAB, where distances match human perception. Colors 2.3 units apart are one "just noticeable difference."
• Role assignment — Colors get roles (dominant, accent, light, dark) based on coverage, contrast, and spatial coherence — not just frequency.
• Salience over coverage — A rare high-contrast accent can outrank a common background color in visual importance.

Installation

Requires Python 3.10+.

# With uv (recommended)
uvx --from git+https://github.com/Capacap/palette-extractor palette-extractor -i image.png

# Or install globally
uv tool install git+https://github.com/Capacap/palette-extractor
palette-extractor -i image.png

Usage

palette-extractor -i <image_path> [options]

Options:

Flag	Description
-i, --input	Path to image file (required)
-o, --output	Write HTML report. Optionally specify path, otherwise auto-names from input.
--no-downscale	Process at full resolution instead of downscaling to 256px

Output

Text output includes:

• Overview — Scheme classification (monochromatic, analogous, complementary, etc.), lightness/chroma ranges, distribution
• Colors — Notable colors ranked by prominence with roles: Dominant, Secondary, Accent, Light, Dark
• Relationships — WCAG-rated contrast pairs and harmonic color pairings

Each color shows hex, RGB, LAB values, coverage percentage, and perceptual notes.

Examples

# Analyze and print to terminal
palette-extractor -i photo.jpg

# Generate HTML report (auto-named photo-palette.html)
palette-extractor -i photo.jpg -o

# Generate HTML report with custom path
palette-extractor -i photo.jpg -o report.html

# Full resolution analysis (slower, more precise)
palette-extractor -i photo.jpg --no-downscale