Use cases
- WeChat Official Account cover(preferred): the user provides the Official Account content → output one cover image.
- Design philosophy + canvas: the user wants a poster/artwork/static visual → first write a visual philosophy (.md), then express it on a canvas (.pdf/.png).
WeChat Official Account cover
- Input: the WeChat Official Account article content provided by the user (or the title, summary, etc.).
- Output: one image, by default saved as
imagen_output.png。
Execution: in the skill directory canvas-design , execute the script, passing the Official Account content entered by the user (either as-is or organized/translated into a description) as the prompt:
python scripts/generate_cover.py "<用户输入的微信公众号内容>"
The script internally calls generate_image_correctly(prompt); when Imagen performs better with English prompts, the copy can first be concisely translated or summarized before being passed in.
Design philosophy + canvas
Two steps: 1) write a visual philosophy and output it as .md; 2) visually express it on a canvas, output as a single-page .pdf or .png (for fonts see canvas-fonts/）。
For the complete workflow, philosophy examples, principles, and refinement steps see reference.md。
Resources
- Script:
scripts/generate_cover.py(for WeChat Official Account covers) - Fonts:
canvas-fonts/ - Dependencies:
requirements-imagen.txt(for cover image generation)