Layerize

Layerize takes a flat graphic image, detects text elements, and returns:

1. A background image with text removed
2. Structured text layer data (JSON) with font information, positioning, colors, and HTML overlay code

The text layer data includes: - Text containers grouped by related items - Font class detection (h1, h2, body, small) - Span-level styling (color, bold, italic, underline, caps) - Precise positioning (x, y, width, height, angle) - Font candidates for each text element - Ready-to-use HTML overlay (text_html)

Use cases

• Edit text in existing designs without recreating the layout
• Extract text layers for translation or localization
• Re-compose graphics with different fonts or colors
• Build design editing tools that understand text structure

Output format

The model returns two files: - An image file (PNG) — the background with text removed - A JSON file — the full API response including text_containers and text_html

Font overrides

You can override fonts per style class using the font_name_* parameters: - font_name_h1 — heading 1 text - font_name_h2 — heading 2 text - font_name_body — body text - font_name_small — small text