director-anime-quality-boost

name

director-anime-quality-boost

description

【导演模式·动画质感 / Director · Anime Look】触发词:动画质感 / 日式动画截图 / 赛璐璐 / cel / 去厚涂 / 油画感 / 二次元 / 番剧风 / anime screenshot / no painterly。Use when output drifts toward painterly / 厚涂 / oil-painterly texture instead of cel-shaded anime screenshot style, character identity must lock across panels (hair / outfit / accessories), or injecting JSON instruction blocks with Color Harmony 1.5 and No-Painterly 1.8 weights (日式动画截图) — applies to image-generation models (Midjourney, DALL-E, FLUX, Stable Diffusion, Imagen, Ideogram, Recraft), video-generation models (Sora, Veo, Runway, Kling, Seedance, Hailuo, Higgsfield, Hunyuan), screenplays, scripts, storyboards, AI video, AI image, 提示词, 视频模型, 图像模型, 写剧本, 脚本, 分镜.

Anime Quality Enforcement — enforces cel-shading, color harmony, character identity lock, and removes painterly/厚涂 texture.

Character Identity Lock (BINDING)

Core principle: 基本人設は変えない、演出は変えてよい (character design stays fixed, performance can evolve).

• Face structure, hairstyle, hair color, outfit design, signature accessories MUST remain recognizable across ALL panels
• Pose, expression, action, lighting on character, camera angle MAY change freely for dramatic effect
• If a character has twin-tails, twin-tails appear in every panel. If a character wears a military tunic, the tunic appears in every panel.
• Reference image is the SINGLE SOURCE OF TRUTH for character identity. Never invent new outfits, hair colors, or accessories.

designAndAssemble Rules

Every panel prompt MUST embed this JSON instruction block BEFORE panel descriptions. The JSON format triggers stronger model attention than plain text:

{
  "instruction": {
    "reference": "Strictly follow the character designs and scene layout from the input image.",
    "style_override": "Transform the visual style into a 1990s-2010s Japanese high-quality anime screenshot style (cel-shaded).",
    "rendering_requirements": [
      "Integrate character skin tones and overall coloring perfectly into the environment's warm lighting and ambient atmosphere (Color Harmony: 1.5).",
      "Completely remove the 'heavy painting' or 'thick oil-painterly' texture from the characters (No painterly texture: 1.8).",
      "Apply clean, sharp line art and flat cel-shading to mimic professional 2D animation frames.",
      "Ensure characters' shadows and highlights are driven by the scene's light sources."
    ],
    "weight_adjustment": {
      "anime_screenshot_style": "+0.99",
      "remove_painterly_texture": "+0.99",
      "lighting_integration": "+0.95"
    }
  }
}

generateImages Rules

Append to composite prompt:

STYLE ENFORCEMENT (BINDING):
Japanese anime screenshot style, cel-shaded, clean sharp line art, flat coloring.
Character tones blend with scene lighting and color palette (Color Harmony: 1.5).
No heavy painting texture, no oil-painterly texture (Weight: 1.8).
Shadows and highlights driven by scene light sources.
CHARACTER IDENTITY LOCK: Strictly follow reference image character designs. Hair, outfit, and accessories MUST NOT change across panels.
将分镜画面风格变为日式动画截图风格
人物色调需融入场景的光感，融入场景色调，去掉人物厚涂质感

Add to negative prompts: heavy painting, thick oil texture, painterly brush strokes, impasto, textured canvas, 3D render, CGI, photorealistic, inconsistent character design, wrong outfit, changed hairstyle

Example

Output drifts toward 厚涂 / oil-painterly → inject the designAndAssemble JSON block BEFORE the panel text, then append the STYLE ENFORCEMENT lines, then add the negative list above. Result: clean sharp line art, flat cel-shading, scene-driven shadows; [char1]'s twin-tails + sailor uniform stay identical across every panel.