Anime Style Prompt Expert
Turn any idea into precise anime sub-style image prompts for today's top models
For illustrators, indie game and comic makers, VTubers, merch sellers and anyone who prompts image models for anime art.
"Anime style" on its own gives you generic results. This skill makes your agent pick a specific sub-style - 90s cel, modern TV key visual, painterly feature film, shoujo, shounen action, real-robot mecha, chibi, watercolor slice-of-life, cyberpunk, ukiyo-e fusion, manga page and more - and describe it through visible traits: line weight, shadow shape, palette, lighting, composition and era artifacts.
It then writes the prompt the way your generator reads it: labeled paragraphs for OpenAI GPT Image 2.5, narrative prompts with aspect ratio and image size for Google's Nano Banana models, concise natural language with size tiers for ByteDance Seedream 5.0, medium-first prompts with bounding-box layouts for FLUX 3 Image, and compact phrases plus parameters for Midjourney V8 and Niji 7. Model facts were checked against official documentation in October 2026.
You get: a decision table for choosing a style, a 20-style trait library with prompt fragments, an anime vocabulary bank (shots, poses, expressions, effects, palettes with hex codes, title lettering), model adapters with parameters and reference-image limits, five worked examples, a self-check list, and a small Python script that converts any ratio or pixel size into valid settings per model.
The skill describes looks instead of naming living artists or copying studio IP and characters, so the images you make stay yours. Results still depend on the model you use; review outputs before publishing.
Highlights
- 20 anime sub-styles described by line, color, shading and light
- Adapters for GPT Image 2.5, Nano Banana, Seedream 5, FLUX 3, Midjourney
- Ratio/size fitting script with valid settings for each model
- Character consistency, sticker sets, covers and manga layouts
- Original-by-design: no artist names, no studio or character copying
Example
Phone wallpaper of my OC Kitsu, a fox-eared courier girl running through a rainy neon city at night. Midjourney please.
**Style:** Cyberpunk anime + wet-reflection accent - neon rim light sells the night run, reflections add depth. **Model:** Midjourney Niji 7 **Settings:** --ar 9:16, moderate stylize **Prompt** cyberpunk anime illustration, sharp clean line art, hard cel shadows split by cyan and magenta neon rim light, a young woman courier with fox ears and a fluffy orange tail, short ash-blond hair, amber eyes, oversized navy bomber jacket with reflective stripes, red messenger bag, sprinting across a rain-slick crosswalk with a parcel under one arm, low angle full body in the lower half, neon buildings and cables above, empty misty sky at the top for the clock, puddles mirroring the signs --niji 7 --ar 9:16 --stylize 150 --no text, letters, watermark **Variations:** A. resting under an awning with a steaming can. B. top-down drone view. **Iterate with:** "Keep everything, make the tail larger".
Retro 90s anime poster: battered mech in a hangar, pilot at its feet, title IRON TIDE at the top, for A3 print.
**Style:** Real-robot mecha + 90s cel finish.
**Model:** FLUX 3 Image **Settings:** aspect_ratio "5:7" (nearest to A3, ~1% crop), resolution "2k"
**Prompt**
A hand-painted 1990s anime cel poster: uniform ink outlines, flat cel colors with one hard-edged shadow tone, gouache-painted hangar background, light film grain. <title> sits across the top. Below it, <mech> towers under harsh sodium floodlights, armor dented and covered in crisp panel lines. At its feet, <pilot> looks up, small for scale. Palette: olive #6B7A4B, off-white #E8E2D0, warning orange #E0662B.
[{"id": "title", "bbox": [30, 60, 170, 940], "desc": "the words \"IRON TIDE\" in heavy condensed italic capitals, off-white with a thin orange outline"}, ...]A girl reading in a library in the style of [a living illustrator].
I described the look you like - glowing window light, fine architectural detail, soft desaturated pastels - instead of naming the artist, so the result stays original. **Style:** Painterly feature film + shoujo accent. **Model:** Seedream 5.0 pro **Settings:** size "2K", horizontal 3:2 stated in the prompt, watermark false if permitted **Prompt** A softly painted anime illustration for a horizontal 3:2 wallpaper. A young woman with long silver-lavender hair sits in a deep window seat of an old library, an open book on her knees. Tall arched windows pour hazy afternoon light across the shelves; dust motes float in the beams. Thin soft line art, simple cel shading, luminous eyes with two catchlights; the library fully painted with fine architectural detail. Palette: dusty rose, sage, warm ivory, pale gold.
What's inside
anime-style-prompt-expert/ ├── agents/ │ └── openai.yaml ├── references/ │ ├── anime-vocabulary.md │ ├── model-adapters.md │ ├── style-library.md │ └── worked-examples.md ├── scripts/ │ └── aspect_fit.py ├── LICENSE.txt ├── README.md └── SKILL.md
Install by unzipping into your agent's skills folder. Install guide →