Ian Xiaohei Weird Body Illustrations
Core Positioning
Design and generate 16:9 landscape body illustrations for Chinese articles. The goal is not to create commercial illustrations, PPT infographics or cute cartoons, but to turn key judgments, processes, structures, states or metaphors in the article into a fresh, weird, creative, readable yet non-manual hand-drawn explanatory image.
The default visual IP is "Xiaohei": solid black, white dot eyes, thin legs, empty expression, seriously doing something absurd but reasonable. Xiaohei must participate in the core action of the image, not just stand aside as decoration.
Read These References First
Read as needed for the task, don't stuff all context at once:
- : Style DNA, colors, text, taboos.
- : Xiaohei IP's image, personality, action library and taboos.
references/composition-patterns.md
references/prompt-template.md
references/qa-checklist.md
- : Only for low-frequency visual calibration, not part of the default generation path. Do not copy the composition, objects or annotations of these cases.
Workflow
1. Digest the Body Text
First read the body text, links, Notion pages, Markdown files or screenshots provided by the user. Extract:
- What is the core viewpoint
- Which paragraphs bear cognitive turning points
- Which content is suitable for illustration explanation
- Which parts are only suitable for text and do not need images
Do not allocate illustrations evenly. Prioritize "cognitive anchors", such as: core judgments, two breakpoints, input-output closed loops, diversion, before-and-after comparison, one fish multiple uses, undertaking paths, common pitfalls, role state changes.
2. Propose Illustration Strategy First
If the user only says "analyze how to illustrate / think about which parts need illustrations", first provide a shot list. For each image, clearly write:
- Which paragraph it is placed after
- Theme of the image
- Core meaning
- Structure type
- What Xiaohei does in the image
- Suggested elements
- Suggested Chinese annotation words
The default is 4-8 images. 1-3 images for very short articles; do not easily exceed 9 images for long articles. Just use enough, avoid turning the body text into an album.
3. Single Image Generation
If the user clearly requests "generate / output / create images / help me generate", do not stop waiting for confirmation; use the built-in
to generate each image separately. Do not combine multiple images into one.
Each image only focuses on one core structure. The prompt must include:
- 16:9 landscape Chinese body illustration
- Pure white background
- Black hand-drawn line draft
- A small amount of red/orange/blue Chinese handwritten annotations
- Large amount of blank space
- Xiaohei as the core action subject
- Prohibit PPT, commercial illustrations, naive cuteness, complex architecture, top-left type titles
Do not replicate past cases. Cases only provide style density and Xiaohei's participation method, and you cannot directly reuse existing compositions such as "conveyor belt breakpoint / Xiaohei pulling line / material fish / stamping toolbox / common pit path" unless the user explicitly requests to replicate a certain image. Invent a new weird but reasonable metaphor from the current article every time.
4. Inspection and Iteration
Check
references/qa-checklist.md
after generation. If the following problems occur, prioritize re-generation or partial editing:
- Xiaohei is only a decoration
- The image is too crowded
- Too similar to flowcharts/PPT
- Too much Chinese or serious typos
- Titles such as "common pit/flowchart/system architecture diagram" appear in the top-left corner
- The style is too cute, naive, rigid
- The background is not a clean white background
5. Save and Deliver
If the user works in the workspace, copy the final images to:
text
assets/<article-slug>-illustrations/
Name them in order:
text
01-topic-name.png
02-topic-name.png
Keep the original generated files, do not overwrite existing assets unless the user explicitly requests replacement.
Output Guidelines
The strategy output before generation should be short and accurate. The delivery after generation should include:
- Number of generated images
- Purpose of each image
- Save path
- Which images are the most stable, which are optional
Do not explain style theories at length; let the images speak for themselves.