Anime Character Prompts: 12 Copy-Ready Templates (2026)

Q: Is `--cref` still the way to keep a character consistent in Midjourney?

No. As of 2026, `--cref` is retired. On V7 and V8.1 you use Omni Reference (`--oref`) with a character weight `--cw` from 0 to 100, which costs about 2x the usual GPU time. Niji 7 has no character reference at all, so for an anime series you rely on `--sref` plus a fixed seed.

Q: How do I stop the model from adding random background objects?

Add `simple background` or `solid pastel background` to the composition line and remove vague atmospheric words. On NovelAI, put a `background` tag at the very start of the prompt.

Q: Best aspect ratio for anime portraits?

`2:3` for character art, `4:5` for social, `16:9` for key visuals, `9:16` for phone wallpaper.

Anime character prompts that produce stable, sellable art. Twelve templates plus the six-layer structure they share, tuned for Niji 7, NovelAI V4.5, and Flux.

Published: May 12, 2026 Updated: Jun 04, 2026 Author: AI Productivity Guide Team 🌐 查看中文版本

Anime character prompts go wrong not because you’re short on keywords, but because the prompt has no structure. Every reliable anime image prompt follows the same six layers: subject + style + composition + lighting + lens + detail constraints. This guide walks through that structure and hands you 12 copy-ready templates covering the most common character archetypes, with model-specific notes current as of June 2026.

TL;DR

Write the prompt in six layers, one per line. Structured prompts beat keyword soup on every modern model.
For anime specifically, Niji 7 (Midjourney’s anime model, launched January 9, 2026) and NovelAI Diffusion V4.5 give the cleanest line art; Flux and Illustrious XL are the strongest open options.
Lock style with a fixed “style header” (the style/composition/lighting lines). Lock the character with a reference image, not just words.
Always set an aspect ratio: --ar 2:3 for character art, --ar 4:5 for social, --ar 16:9 for key visuals, --ar 9:16 for wallpaper.

Which model for anime in June 2026

Tool	Best model / version	Character consistency	Notes
Midjourney	Niji 7 (anime) / V7 default, V8.1	`--oref` + `--cw 0–100`; `--sref` for style	`--cref` is retired; Niji 7 has the cleanest linework for anime
NovelAI	Diffusion V4.5 Full/Curated	Built-in multi-character prompting	Tag-driven; ~512 T5 tokens; plans $10 / $15 / $25 per month
Flux	FLUX.1 Kontext (Max/Pro/Dev)	Character carry-over across edits	Strong at photo-to-anime style transfer
Local SD	Illustrious XL / NoobAI XL / Animagine XL 4.0 / Pony V6	ControlNet + IP-Adapter / LoRA	”Division of labor” — pick the model per style

One reframe worth internalizing: in 2026, consistency comes from reference images and seeds, not from longer text. Midjourney’s old --cref is gone; on V7/V8.1 you use Omni Reference (--oref) with a character-weight value --cw from 0 to 100. Niji 7 does not support character reference at all, so for an anime series you lock the look with --sref plus a fixed seed.

What a good anime prompt actually contains

Whether you use Niji 7, NovelAI V4.5, Flux, or Stable Diffusion, these layers all need to be present:

Subject: gender, age, hairstyle, eye color, clothing, accessories
Style: anime style, cel shading, watercolor anime, Ghibli style, Type-Moon style, etc.
Composition: upper body, half body, full body, close-up, from above
Lighting: soft rim light, golden hour, neon backlight, moonlight
Lens: 85mm portrait, 35mm wide angle, shallow depth of field
Mood and detail: color tone, background elements, emotional vibe
Aspect ratio: --ar 2:3 for portrait, --ar 9:16 for phone wallpaper

Hidden trick: write each layer on its own line instead of mashing them into a single sentence. Models parse structured prompts far more reliably.

12 copy-ready prompt templates

1. Cool school boy (academic vibe)

anime portrait of a slender 17-year-old boy with messy silver hair and pale blue eyes, wearing a dark navy school uniform with loose tie, neutral expression, looking slightly away
Style: modern cel shading, clean line art, soft gradient shading, A-1 Pictures style
Composition: upper body, slightly off-center, blurred classroom in background
Lighting: cool morning window light from the left, soft rim light on hair
Lens: 85mm portrait, shallow depth of field, --ar 2:3

Best for character sheets, light novel covers, social avatars. Swap silver for black or ash blonde to change the vibe.

2. Sweet girl (healing style)

anime portrait of a cheerful 16-year-old girl with shoulder-length light brown hair tied in a low ponytail, warm hazel eyes, soft smile, wearing a cream knit sweater and beige skirt
Style: Kyoto Animation style, gentle painterly shading, pastel color palette
Composition: half body, three-quarter view, holding a paper coffee cup
Lighting: soft golden hour sunlight from window, warm bokeh background
Lens: 50mm f/1.8, shallow depth of field, --ar 2:3

Tuning: swap Kyoto Animation style for Makoto Shinkai style to get a higher-saturation, more cinematic light.

3. Cyberpunk character

anime portrait of a 22-year-old female hacker with platinum white hair shaved on one side, sharp violet eyes, faint scar across cheek, wearing a black techwear jacket with glowing cyan circuit patterns
Style: modern anime key visual, cyberpunk anime style, vibrant color contrast
Composition: half body, slight low angle, neon-lit alley background
Lighting: magenta and cyan neon rim light, wet pavement reflection
Lens: 35mm, shallow depth of field, --ar 16:9

The techwear keyword is doing real work here — it pushes the model toward modern, structured clothing instead of generic sci-fi armor.

4. Chinese aesthetic anime (Xianxia warrior)

anime portrait of a young swordsman in flowing white robes with red trim, long black hair tied half-up, calm focused expression, ornate silver sword at his side
Style: modern Chinese animation, semi-3D shading, ink-wash background elements
Composition: half body, three-quarter view, mountains in mist behind
Lighting: soft morning light from above, gentle wind motion on robes
Lens: 50mm portrait, --ar 2:3

Guofeng keywords that pull their weight: hanfu, ink-wash background, Mihoyo-inspired, silver embroidery. A crimson + black + jade palette usually reads better than a flat “traditional red.”

5. Battle scene character

anime action portrait of a 20-year-old warrior mid-leap with twin daggers, sleeveless dark crimson combat outfit, intense focused glare
Style: vibrant anime key visual, MAPPA-style dynamic shading, bold line art
Composition: full body, dynamic diagonal pose, motion blur on weapons
Lighting: dramatic backlight from explosion, embers in air
Lens: 35mm wide angle, --ar 16:9

What makes the action read as motion: mid-leap + dynamic diagonal pose + low angle, stacked together. Drop any one and the figure goes static.

6. Magical girl

anime portrait of a 14-year-old magical girl with twin-tail pink hair and large amethyst eyes, wearing a ruffled white-and-pink dress with star accents, magical wand raised, sparkles around
Style: classic shoujo anime, soft cel shading, pastel rainbow palette
Composition: full body, slightly low angle, sparkling background
Lighting: warm key light, soft pink rim, magical glow effects
Lens: 50mm, --ar 2:3

The non-negotiable trio for this look: saturated color, sparkles/glitter, and ribbons. Miss one and it stops reading as shoujo.

7. Chibi character

Chibi anime character, big head small body proportion, cute round eyes, wearing a hoodie with cat ears, plain pastel background, kawaii sticker design, clean vector look, --ar 1:1

Best for stickers, mascots, app icons. Add die-cut sticker, white border if you want it print-ready.

8. Anime key visual (movie poster style)

Anime movie poster, two heroes standing back-to-back on a rooftop at sunset, dramatic wind in coats, vibrant blue and orange complementary palette, modern anime key visual, ultra-clean line art, cinematic composition with title space at top, --ar 2:3

The title space at top cue reserves negative space so you can drop in a logo later without recropping.

9. Shinkai-style high schooler

A 17-year-old Japanese high school girl in summer uniform, soft wind in her hair, looking up at sky, Makoto Shinkai style, hyper-saturated cinematic anime key visual, golden hour clouds, lens flare, ultra-detailed background, --ar 16:9

Shinkai’s signature is the sky, not the face — keep golden hour clouds, lens flare, and ultra-detailed background even when you crop tight.

10. 90s retro anime aesthetic

anime portrait of a 19-year-old woman with short bob cut and red lipstick, wearing oversized denim jacket and yellow scarf, 1990s anime aesthetic, retro cel shading, slightly grainy film look, muted warm color palette, --ar 4:5

slightly grainy film look + muted warm color palette are what sell the era. Skip them and you get a modern face in vintage clothes.

11. Dark fantasy mage

anime portrait of a 30-year-old male mage with long silver-grey hair, intense violet eyes, faint scar across left brow, wearing a black hooded robe with gold runic embroidery
Style: dark fantasy anime, painterly shading, muted desaturated palette with gold accents
Composition: half body, slight low angle, ruined cathedral in background
Lighting: moody side light, gold rune glow on face
Lens: 85mm portrait, shallow depth of field, --ar 2:3

Swap mage for fallen knight or exiled prince and keep the rest — the lighting and palette carry the dark-fantasy mood regardless of the role.

12. Slice-of-life classroom

A teenage girl reading a book by a sunny classroom window, Kyoto Animation slice-of-life style, soft daylight, pastel color palette, gentle warm shadows, cel-shading, --ar 4:5

The looking back over shoulder variant plus late afternoon golden light is the classic “that feeling” frame — try it as a swap for the reading pose.

How to keep the style consistent across a series

If you’re producing a multi-image series, words alone won’t hold a character together. Combine these:

Lock the seed: Midjourney --seed 12345; every SD/Flux UI lets you set a fixed seed.
Reuse a fixed “style header”: copy the style/composition/lighting lines verbatim, only change the subject.
Use a reference image:
- Midjourney V7/V8.1: Omni Reference --oref <url> --cw 60 (--cw 100 locks face and clothing, lower values free up styling). Note --oref costs roughly 2x the normal GPU time.
- Niji 7: no character reference — lean on --sref <code> --sv 6 plus a fixed seed.
- NovelAI V4.5: use its native multi-character prompt blocks.
- Stable Diffusion: ControlNet + IP-Adapter, or train a small character LoRA.
Limit the color palette: palette: warm cream, faded teal, dusty pink.

Common mistakes

Prompt too generic (anime girl, pretty, cute) — the model has no anchor.
Style words conflict (Shinkai style + 90s cel-shaded retro → muddy output). Lock one style per image.
“Magic words” overload: masterpiece, best quality, 8k, stunning do almost nothing on current models and crowd out the descriptions that matter.
No lens specified — the model defaults to a tight face crop with random depth of field.
No aspect ratio — defaults to square and crops awkwardly.

FAQ

Q: Is --cref still the way to keep a character consistent in Midjourney? A: No. As of 2026, --cref is retired. On V7 and V8.1 you use Omni Reference (--oref) with a character weight --cw from 0 to 100, which costs about 2x the usual GPU time. Niji 7 has no character reference at all, so for an anime series you rely on --sref plus a fixed seed.

Q: Niji 7 or NovelAI V4.5 for anime? A: Niji 7 (launched January 9, 2026) gives the cleanest, most production-like linework and is easiest if you’re already in Midjourney. NovelAI Diffusion V4.5 wins for tag-precise, multi-character scenes and gives you more control over explicit composition; plans run $10 / $15 / $25 per month as of June 2026.

Q: Should I prompt in English even on a Chinese SD UI? A: Yes. Most diffusion models were trained on English captions, so style, lens, and lighting terms are most stable in English. Some Chinese models (Kling, Jimeng) handle Chinese well, but the “industry” vocabulary is still best in English.

Q: Why do my faces still look “AI weird” even with these templates? A: Usually it’s the aspect ratio (extreme ratios warp faces) or an outdated base model. On Midjourney move to Niji 7 / V8.1; on local SD switch to Illustrious XL or NoobAI XL, which fix hands and anatomy better than older anime checkpoints.

Q: How do I stop the model from adding random background objects? A: Add simple background or solid pastel background to the composition line and remove vague atmospheric words. On NovelAI, put a background tag at the very start of the prompt.

Q: Best aspect ratio for anime portraits? A: 2:3 for character art, 4:5 for social, 16:9 for key visuals, 9:16 for phone wallpaper.

Tags: #Anime #Image generation #Prompt #Portrait