Bornforthis Illustration Skills

Bornforthis Crayon Hand-Drawn Inline Illustrations

Core Positioning

Design and generate 16:9 horizontal inline illustrations for Chinese content. Turn key judgments, processes, structures, states, or metaphors into a clean, creative, readable, non-instructional hand-drawn explanatory image. Do not make them into commercial illustrations, PPT infographics, or course slideware.

The default visual IP is Bornforthis: a young male character with cobalt blue short spiky hair, outward-round ears, a paper-white round-square face, yellow round cheeks, and minimal blue facial features. Bornforthis must participate in the core action of the image, not just stand nearby as a sticker.

It is the character identity that is fixed, not the outfit. Clothes, headphones, shoes, bags, cameras, books, and other accessories or props must be re-selected according to the current topic and must not default to those in reference images.

Reading References

Read the corresponding files based on the task stage:

• Before any generation or character redraw, first read references/bornforthis-ip.md and references/style-dna.md.
• When planning illustrations and inventing metaphors, read references/composition-patterns.md.
• When writing prompts for generation or edits, read references/prompt-template.md.
• After generating each image, read references/qa-checklist.md and check every item.
• When visual calibration is needed, first read references/example-index.md, then select at most 1–2 images from assets/examples/ based on the issue; do not load the entire set at once.

assets/examples/ contains 14 finished Bornforthis reference images, covering: two breakpoints, sorting by purpose, information compression, causal chain, abstract to concrete, contrast differences, process steps, core anchor, note backflow, information well, creative press, content fermentation, system bearing, and trust bridge. They only calibrate the white background, texture, strokes, whitespace, character consistency, outfit range, and character participation — they never serve as default composition templates.

Each new generation must pass both of the following identity reference images in a single image_gen call; do not pass only one:

• assets/references/bornforthis-ref-1.png: face and head identity reference.
• assets/references/bornforthis-ref-2.png: full body proportion and pose reference.

Read only the fixed face, hairstyle, ears, facial features, proportions, and hand-drawn texture from the two reference images. Do not copy the yellow background, white sticker edge, headphones, clothing, shoes, camera, bag, book, pose, or surrounding symbols from them.

Distinguish between two categories of built-in images:

• assets/references/ are identity anchors: must pass both in every generation.
• assets/examples/ are finished visual examples: only check when calibration is needed; do not pass them to image_gen by default. If text specification conflicts with examples, the text specification and the two identity anchors take precedence.

Workflow

1. Digest the Source Text

First read the user-provided text, links, documents, or screenshots, and extract:

• The core point.
• Paragraphs that carry cognitive shifts.
• Cognitive anchors suitable for illustration.
• Content that is only suitable for text and does not need images.

Do not illustrate evenly. Prioritize core judgments, breakpoints, input-output cycles, branching, before-and-after comparisons, connecting paths, common pitfalls, and character state changes.

2. Output the Illustration Strategy

If the user only asks to "analyze how to illustrate / think about which parts need images", first provide a shot list. For each image, clearly describe:

• Which paragraph it follows.
• The image's theme and core meaning.
• The structure type.
• What Bornforthis is doing.
• Suggested elements and Chinese annotation words.

Default: 4–8 images; for short text: 1–3 images; for long text, do not exceed 9 images easily.

3. Single Image Generation

If the user explicitly asks to "generate / output / make an image / help me generate", do not stop to wait for confirmation. Call image_gen individually for each image; do not combine them into a group image.

Enforce these hard rules:

1. Place the full identity description of Bornforthis at the very front of the prompt.
2. Pass both fixed reference images; declare them as "identity references, not clothing or composition references".
3. Design the clothing and accessories for the current drawing based on the topic; do not automatically add headphones, cameras, books, or outfits from reference images if there is no semantic need.
4. Each image expresses only one core structure. Prefer to simplify the background rather than sacrifice character identity precision.
5. Keep 16:9, pure white background, dark blue hand-drawn thick strokes, visible crayon diagonal texture on all color patches, a small number of blue and red Chinese short annotations, and plenty of whitespace.
6. Do not replicate poses from reference images or past cases; invent a strange but valid metaphor from the current source text.

4. Check and Iterate

Check against references/qa-checklist.md. If identity drift, incorrectly fixed clothing, solid flat color, yellow background, too-filling composition, PPT-like feel, too many Chinese characters or typos, or top-left type title appears, prefer local editing or regeneration.

When editing only involves removing text or cleaning the background, protect character and composition unchanged. When editing will redraw Bornforthis, pass the target image together with the two fixed identity reference images.

5. Save and Deliver

When working inside a workspace, save the final images to:

assets/<article-slug>-illustrations/

Name them in order:

01-topic-name.png
02-topic-name.png

Keep the original generated files, do not overwrite existing assets, unless the user explicitly asks for replacement.

Output Tone

Pre-generation strategy must be short and precise. Post-generation delivery includes: number of images generated, each image's purpose, save path, and which images are the most stable and which are optional. Do not give long-winded explanations of style theory.