Create pixel art with AI
Author compact JSON drawings with named colours, layers and reusable shapes, then use the separate local toolkit to render a PNG.
Create pixel art with AI by describing small images as editable JSON rather than long lists of pixels. The pixel-art-json skill favours named palette colours, layered drawing operations and reusable components for repeated details.
The skill itself returns JSON for the Pixel Art Skill Toolkit renderer. A separately prepared local Python and Pillow renderer turns that source into a PNG; the optional desktop editor also needs Tkinter. Rendering does not require an image-service account or API key, although your chosen agent receives the task and source files. Integer scaling preserves hard pixel edges, while changes can remain limited to one colour, operation or layer. Keep the JSON alongside the export so the drawing remains understandable and easy to revise.
What it does
- Choose a small canvas and restrained colours for an icon.
- Keep colour edits in a named palette.
- Reuse one component for repeated windows or eyes.
- Revise one layer while preserving the rest.
- Inspect the PNG at native and enlarged sizes.
Install and use
- Load the guide and prepare the toolkit’s renderer, then specify canvas dimensions, subject and colours.
- For example: “Make a 24-by-24 three-colour plant icon with a transparent background and reusable leaves.”
- Validate and render the JSON, inspect the PNG and request a focused change to one palette entry or layer.
FAQ
Source
Made by Pixel Art Skill Toolkit contributors. Original skill: pixel-art-json, published on GitHub under the MIT licence.
View on GitHub