This lesson addresses the common problem of AI images that are technically acceptable but visually weak. Instead of asking for a painter imitation, it turns visual judgment into parameters and tests three iterations.
The final Skill preserves the recognizable structure and colors of a source photo while radically simplifying the scene, adding generous negative space, and keeping a loose hand-drawn line.
Make taste executable
Control information density, composition, sampled color, line quality, texture, and forbidden elements. Keep only two to four key shapes, remove most detail, use three to five muted colors, leave 65–80% negative space, and change only one to three variables per iteration.
The full Chinese source prompts remain on the source-language page; this English note captures the operational rule so the visual Skill can be reviewed and reused.
Transform the uploaded photo into a restrained contemporary editorial sketch for a presentation. Preserve its most recognizable silhouette and sampled color accents, but keep only 2–4 core shapes, remove over 90% of details, use 3–5 muted colors, leave 65–80% negative space, and use loose imperfect hand-drawn lines. Avoid photorealism, 3D, heavy ink wash, glossy vector style, clutter, and subjects touching the edges.