<prompt version="3.0">
<meta>
Interpret all directives as binding constraints.
"Redraw the entire image from scratch while incorporating all requirements."
Do not cut, paste, collage, or photobash from references; re-synthesize the whole image as a clean 2D illustration.
</meta>
<priority>
1=reference identity & exact style match
2=pose & coffee action
3=viewpoint & framing
4=scene/background (café interior)
5=style/lighting coherence
</priority>
<reference>
Image #1: the uploaded illustration of an anthropomorphic tabby cat in a black suit with red tie and red pocket square.
Treat Image #1 as authoritative for: species (cat), head/ear shape, whiskers, tabby stripe pattern (forehead & ringed tail),
facial proportions/eyes/muzzle, suit cut and palette, line weight, flat-shape rendering, and overall mid-century cartoon style.
</reference>
<rendering>
Redraw from scratch in the same 2D mid-century cartoon style as Image #1:
clean, confident linework; flat shapes with restrained shading; limited, harmonious palette; subtle paper-like texture allowed.
Avoid photoreal/PBR cues, heavy gradients, or noisy textures.
</rendering>
<character>
<identity>
Same character as Image #1: anthropomorphic tabby cat with upright triangular ears, short whiskers (3 per side),
forehead stripes, ringed tail; black single-breasted suit, white shirt, red tie, red pocket square, brown dress shoes.
Maintain the original facial attitude (subtly confident).
</identity>
<pose>
Seated elegantly on a café chair at a small table.
Right hand holds a porcelain coffee cup near the mouth; left hand lightly supports the matching saucer on the table or under the cup.
Gentle, poised sip (“sipping” in action), shoulders relaxed, spine tall; legs together and composed (no slouching).
Expression: subtle closed-mouth smile (no teeth); eyelids slightly lowered, friendly and self-assured.
Tail forms a soft S-curve behind or beside the chair; no awkward intersections with furniture.
</pose>
<accessories>
Keep tie and pocket square exactly red and neatly arranged; suit silhouette and seam placement unchanged.
Preserve whisker count/length and tabby markings; do not add or remove stripes.
</accessories>
</character>
<props>
<coffee-service>
White porcelain cup and saucer; thin rim; light steam visible; coffee surface calm (no splash).
Optional small spoon on saucer; no brand logos or readable text.
</coffee-service>
<furniture>
Simple café round table (wood top) and chair with clean geometry; proportions normal; no ornate distractions.
</furniture>
</props>
<scene>
Cozy café interior with warm materials (wood, soft wall tones).
Background elements simplified: counter/shelves/plants as abstracted shapes; no other patrons or staff in frame.
</scene>
<background>
Mild stylized depth (soft painterly falloff), not photographic bokeh.
Keep composition tidy around the character; avoid clutter, signage, or readable menus.
</background>
<viewpoint>
Eye-level, front three-quarter view.
Framing includes the full seated figure from ears to shoes and the tabletop with cup/saucer; avoid cropping ears, hands, tail tip, or cup.
Lens plane parallel to the ground (no roll).
</viewpoint>
<style>
Flat shading with minimal highlights; shadow shapes crisp and graphic.
Line weight consistent with Image #1; palette restrained (black suit, red accents, warm café neutrals).
</style>
<lighting>
Warm indoor café lighting: soft key from front-side, gentle fill from room bounce.
Subtle shadow under cup/saucer and on tabletop; no harsh bloom or lens effects.
</lighting>
<mood>
Elegant, composed, and lightly playful; a refined coffee pause.
</mood>
<assert>
identity-style-lock: species, facial proportions, tabby pattern, suit/tie/pocket-square match Image #1.
coffee-sip-lock: cup near mouth, sipping action readable; steam present.
closed-mouth-smile: lips closed; no teeth shown.
full-figure-framing: ears, hands, cup, and shoes fully inside frame; tail tip not cropped.
single-subject-only: exactly one character in frame; no other people.
</assert>
<negative>
no photoreal/PBR/glossy materials; no heavy gradients/noise;
no extra people/silhouettes; no brand logos/readable text;
no anatomy/pose slouching; no cropping of ears/hands/tail tip/cup; no watermark.
</negative>
</prompt>
결과물
대충 이런 식임
1줄 요약 : GPT-5 띵킹 붙잡고 XML로 이미지 모델 프롬프트 세분화시켜서 짜달라고 문답하면 저런거 나옴
걍 대충 해줘 하면 되긴해
XML이나 JSON은 AI들이 잘 알아먹는 만큼 잘 만들어줘서, AI한테 해줘 하면 잘 나오는 듯