Compilea theme or plotintoaprompt that can be handed directly to an image model. The style is locked: traditional Chinese lianhuanhua (little picture book) style single-line baimiao, pure white paper surface, pure black ink lines, strict ninety-degree side view, abstract symbolic expressions, with figures carrying a plain, humorous, slightly exaggerated marketplace quality.
Core principles:The theme only changes the plot, not the drawing method. The entire recognizability of this style comes from four things—baimiao lines made in one pass, symbolized side profiles, real contact, and pure black and white. Any urge to turn lines into sketch hatching, turn side profiles into front views, turn interaction into side-by-side placement, or let gray tones creep into the image must be suppressed.
Aspect ratio
Default 3:4 vertical, image output parameter 1536×2048、quality high. Follow other ratios when the user specifies them; this is the only item in the Composition base that can be changed:
| Scene | Ratio | size |
|---|---|---|
| Article/social media illustration (default) | 3:4 | 1536×2048 |
| Square social media card | 1:1 | 2048×2048 |
| Horizontal embedded | 4:3 | 2048×1536 |
For horizontally extended two-person interaction, rope pulling, carrying, or long props, the figures can be appropriately reduced, but the action relationship must be clear and the surrounding white space must be ample—these are already locked in the Composition base, so no extra writing is needed.
Locked style core
No matter what content is drawn, the following four sections go verbatim into every prompt, without rewriting or condensing.When producing as a group, every prompt must repeat them—the model can only see a single prompt, and shared rules are not automatically inherited. Only the aspect ratio item at the beginning of the Composition base changes with the ratio setting.
Scene base
flat opaque pure white paper background filling all four corners of the frame, completely blank; no environment, no architecture, no furniture, no landscape, no decorative patterns, no background figures; if the action needs context, keep only one or two instantly recognizable props directly tied to the core action, nothing else; five to eight short, loose black ground strokes of slightly uneven length directly under the figures' feet to suggest the ground.
Key details base
traditional Chinese lianhuanhua (pocket comic-book) style, single-line ink drawing; figures built from a few large continuous contour lines, with slightly oversized heads, simplified torsos and simplified limbs; each line drawn like one confident pass of a black brush or dip pen — outer silhouettes use medium-thick, clear, forceful black lines, while facial features, clothing folds and prop interiors use slightly thinner lines; lines are fluid and economical with a slight natural wobble and hand-drawn irregularity, never re-traced, never sketchy search lines, never smooth mechanical vector curves; edges stay sharp and crisp with no softening, no ink bleeding, no texture erosion; strict black-and-white: pure black ink lines, a few solid black fills and the pure white paper — no grey, no colour, no gradients, no shading, no hatching, no cross-hatching, no screentone dots, no grain, no paper texture, no brush noise, no ageing effects; faces are highly abstract and symbolic — the profile eye is a single short arc or a small black dot, the eyebrow a single short arc, the nose a simple hooked profile line, the mouth one short curve or a simple black-and-white shape, the ear carries only one or two summary lines inside; no realistic eyeballs, no eye bags, no teeth, no gums, no heavy wrinkles, no nostril detail, no skin texture, no sweat drops, no facial muscles; clothing is loose, generalized, mostly white, with at most one to three fold lines per sleeve, torso or trouser leg that only explain the direction of the action; no fabric realism, no dense cloth wrinkles, no stitching, no lighting, no volume on clothes; hands are simplified lianhuanhua symbols — the palm reads as one shape, fingers as a few short arcs that read the grip, support, push, pull, point or carry of the action; no fingernails, no knuckles, no palm lines, no tendons, no fingers fusing together, no palm floating next to a prop; solid black fills reserved for hair, shoes, the inside of an open mouth or one key visual anchor, while clothing and large props stay mostly white; mood: naive, humorous, mildly exaggerated, everyday-life folk satire.
Composition base
3:4 vertical; flat eye-level side observation, no bird's-eye view, no worm's-eye view, no wide-angle distortion, no dramatic perspective, no strong foreground-background scaling; the main figure or interacting group centered, occupying about 65% to 85% of the frame width and 75% to 88% of the frame height, with extra white space in front of the figure in the direction of the action; full-body framing — head, both hands, key props, contact points and both feet fully inside the frame, never cropped at wrists, fingers, elbows, knees, ankles or mid-prop; for wide two-figure interactions, rope pulling, carrying or long props the figures may sit slightly smaller, but the action relationship stays instantly readable and the white margins around all four sides stay generous.
Constraints base
no text, letters, numbers or symbols anywhere in the image; no modern western caricature grimaces, no realistic portraits, no real human anatomy, no muscle lines, no detailed facial anatomy, no dense wrinkles, no sweat drops, no explosion symbols, no dense action lines; no frontal or three-quarter faces — strict 90-degree side profiles only, each face shows exactly one eye, one eyebrow and one ear, never a second eye, a second brow or the other ear; no photographs, no 3D rendering, no anime, no manga, no western superhero comics, no fashion illustration, no smooth vector art, no pencil shading, no watercolour, no oil painting, no realistic sketching; no greyscale shading, no volume lighting, no gradients, no screentones, no hatching, no cross-hatching, no paper grain, no scan noise, no ageing texture; no dense interior detailing, no repeated outlines, no mechanical prop details that clash with the figure's line style; no complex backgrounds, no architecture, no furniture arrangements, no landscapes, no decorative borders, no titles, no captions, no speech bubbles, no text labels, no signatures, no seals, no logos, no watermarks; no transparent background — the background must be fully opaque, even pure white; no unrelated figures, no extra props, no extra arms, no extra hands, no extra fingers, no fused limbs, no floating props, no broken ropes, no ambiguous contact.
Plot translation (turning the theme into an image)
First judge the theme as a type of plot, then decide the character configuration. The theme only provides character identity, action, props, and plot; the style is fully taken over by the bases above.
| Plot type | Image configuration | Key points |
|---|---|---|
| Single-person action / labor | One person in side view, facing the direction of action, one prop | The core action is understood at a glance |
| Single-person state | One person in side view, expressing the state through silhouette and body tilt | The state relies on posture, not grimaces |
| Two-person handoff / dialogue | Two people facing each other, with a handed-off object or gesture in between, and the contact point drawn solid | Both ends of the handed-off object must be touched by hands |
| Two people sharing a burden / traveling together | Two people facing the same direction, jointly carrying, lifting, or pulling one prop | All four hands are on the prop |
| Two-person confrontation / pulling | Two people facing each other, with a rope or object in the center, and hand contact at both ends | The rope must be taut, the hand must grip it |
| Supporting / Embracing | Two people in physical contact, hands on shoulders, back, or arms | Body contours interlock, contact points are clear |
| Person and animal | One person plus one animal, with a line-of-sight or leading relationship | Draw the contact points of the leash, feeding, or petting as solid |
Facing rules: A single person faces freely, determined by the direction of the action; two people either face the same direction (shared goal) or face each other (handover, conversation, confrontation, pulling), determined by the narrative relationship. No matter which way they face, each face has only one eye, one eyebrow, one ear, and a side-view nose—front-facing faces and three-quarter faces are never allowed.
Silhouette personality: A character's personality must first be understood through silhouette and pose. Choose one from the silhouettes; do not pursue realistic human proportions:
| Personality | Silhouette treatment |
|---|---|
| Round and plump | Arced outer contour, compressed head-to-body ratio, protruding belly curve |
| Tall and thin | Vertical elongated contour, slender limbs |
| Hunchbacked | Large forward-curving arc of the spine, head jutting forward |
| Upright | Vertical axis, level shoulders |
| Short | Overall compressed, short limbs |
Silhouette test: After filling the figure in as a pure black silhouette, the core action should still be recognizable. Run through this test in your mind before writing the Subject.
Iron rule of contact points: When characters are handing something over, supporting each other, embracing, pushing and pulling, fighting over something, or carrying something together, the relationship must be drawn—hands, rope, water flow, basket, tool, or other connecting object must make accurate contact; do not simply place the characters and objects side by side. The Subject paragraph must specify "which hand, touching which thing, where the contact point is," for example both hands grip the carrying pole in front of and behind the shoulder, the pole resting exactly on the shoulder。
Props: At most two, recognizable at a glance by name, keeping only the outer contour and key structures needed for recognition; complex props follow the simplification level of the character line art, without mechanical textures or dense structural lines.
Black blocks: Pure solid black blocks are used only for hair, shoes, the inside of an open mouth, or one key visual focal point. Name who gets it in each image.
Input
Infer from user input; filling out a form is not required:
- Theme: Character identity, action, props, or plot. One sentence is enough.
- Ratio: Default 3:4, change only if the user specifies.
- Single / Group: For a group, first list clearly which segment each image covers, to avoid two images covering the same thing.
Workflow
- Read the theme → Determine the plot type (single-person action / two-person interaction / person and animal), and consult the plot translation table.
- Define the core action: State this action clearly in one sentence; it must be understandable at a glance.
- Define character configuration and facing: One to two main characters, adding an animal or one key prop only when the theme truly requires it; facing is determined by the plot.
- Define silhouette personality: Choose one from round and plump / tall and thin / hunchbacked / upright / short, and do the silhouette test.
- Define contact points: Write the contact positions of hand to hand, hand to prop, and connecting objects in specific terms.
- Define props and black blocks: Props ≤ two items, nameable at a glance; point out the black blocks to give hair / shoes / inside of mouth / center of gravity.
- Compile by the seven slots：
Use case → Scene → Subject → Key details → Composition → Text in image → Constraints, copy the four-segment base verbatim, and write only the Subject. For the iron rules and skeleton seereferences/compile-contract.md。 - Output: decision summary + one complete prompt + one sentence of translation rationale. When the user asks only for the prompt, omit the explanation.
Output Format
## 已定决策
- 情节类型：
- 人物配置与朝向：
- 轮廓性格：
- 接触点：
- 道具：
- 黑块：
- 比例：
## 完整提示词
```text
...
```
## 转译理由
...
Quality Gates
- Strict profile: each face has one eye, one eyebrow, one ear; no frontal view, no three-quarter view. If the profile shows a second eye or the other eyebrow, it fails.
- Readable silhouette: after the figure is filled in black, the core action is still recognizable. If the silhouette does not read as the action, change the pose first, then write.
- Pure black and white: black lines, black blocks, white paper; no gray, no color, no gradients, no halftone dots, no hatching, no grain, no paper texture. If the output looks gray, it fails.
- Opaque white background: the background is opaque, even, pure white; all four corners are pure white; no transparent areas.
- Contact is real: hands touch the props, the rope does not break, the carrying pole does not float. Placing the figure and the object side by side is as good as not drawing the relationship.
- Full body in frame: head, both hands, key props, contact points, and both feet are fully in frame; no cropping at the wrist, fingers, elbow, knee, ankle, or in the middle of a prop.
- Zero text in the image: no title, caption, speech bubble, label, signature, seal, logo, or watermark.
- One-pass lines: no repeated outlines, no sketch search lines, no smooth vector curves.
- Style bans: no photo, 3D, anime, Japanese manga, American comics, fashion illustration, vector, pencil sketch, watercolor, oil painting, or realistic sketch.
- This skill only produces prompts; it does not fabricate already-generated image results.
Read
- Compilation contract (iron rules + finished-piece skeleton):
references/compile-contract.md - Translation examples (decision table + finished prompt):
references/examples.md