Description
The usual problem with article covers is drawing each keyword as an icon and ending up with a template. space-image is an image-generation skill that first turns a long text into relationships and then ties them together with one central metaphor in a concept poster, with no icon pile.
The default output is 2.35:1 landscape at about 1920×817, the proportion of a WeChat article header.
Extract, then draw: Pulls out topic, keywords, relation sentences, central metaphor and visual conflict before any drawing.
Nine composition grammars: Black chrome, minimal mascot, streetwear character, hand-drawn lifestyle, monochrome print, fluorescent black, cobalt riso, acid xerox collage and spot-colour engraved portrait, matched to the content's mood rather than one layout in different colours.
Restrained text: Only a title and 2 to 4 short labels, large type used as shape, on a solid or near-solid background.
Xerox sub-skill: A separately installable skill for black, blue and pink xerox-collage editorial covers.
It is a set of prompts and rules, so the actual image comes from the generation model you connect. The repository states no license.
The default output is 2.35:1 landscape at about 1920×817, the proportion of a WeChat article header.
Features
Extract, then draw: Pulls out topic, keywords, relation sentences, central metaphor and visual conflict before any drawing.
Nine composition grammars: Black chrome, minimal mascot, streetwear character, hand-drawn lifestyle, monochrome print, fluorescent black, cobalt riso, acid xerox collage and spot-colour engraved portrait, matched to the content's mood rather than one layout in different colours.
Restrained text: Only a title and 2 to 4 short labels, large type used as shape, on a solid or near-solid background.
Xerox sub-skill: A separately installable skill for black, blue and pink xerox-collage editorial covers.
It is a set of prompts and rules, so the actual image comes from the generation model you connect. The repository states no license.
