30+ Best Anime AI Prompts: Midjourney & Stable Diffusion (2026)

Anime is the most requested style in AI image generation — and the one where copied prompts fail most often, because most of them lean on franchise and character names. Models imitate those badly, and you can’t publish the results anyway. Every prompt here describes the look instead: cel shading, hand-painted film backgrounds, retro OVA grain, modern key visuals.

Anime runs 18 prompts deep in our 200-prompt tested library, and every SDXL entry among them ships a negative prompt — as do 100% of our 48 SDXL prompts overall — because cel-clean lineart dies without one (full corpus data).

The model split matters more here than anywhere. Midjourney’s Niji mode is anime-tuned out of the box — describe composition and mood in prose and it handles the style. Stable Diffusion XL demands setup (a negative prompt is mandatory) but repays it with control: fixed seeds, turnaround sheets, anime-tuned checkpoints. We tested every prompt on the model it’s written for — 31 in total, 16 Midjourney and 15 SDXL. The embedded cards carry full settings and usage notes; the collection keeps growing in our anime style hub.

How anime prompts differ from other AI prompts

Production language beats art language. Anime models respond to animation-pipeline vocabulary: “cel shading”, “key visual”, “anime film still”, “painted background”. The highest-leverage phrase is an era anchor — ”90s cel animation”, ”80s OVA with film grain”, “modern anime film still” — because each era is a coherent look the model has seen thousands of times.

Franchise and character names are a trap twice over. The copyright problem is obvious; the quality trap is less so — naming a show gets you a blurry average of ten thousand fan images. Describing the aesthetic instead (flat cel shading, nostalgic muted colors, hand-painted clouds) produces cleaner images that are actually yours.

SDXL needs a negative prompt, always. Anime linework amplifies the anatomy errors soft photorealism would hide. Our baseline for every SDXL prompt in this post:

photorealistic, 3d render, western cartoon, extra fingers, deformed hands, extra limbs, blurry, text, watermark

Midjourney puts exclusions behind a --no flag instead — our guide to writing Midjourney prompts covers how the flags fit together.

Anime street scenes and landscape backgrounds

Scenery is the highest-percentage anime genre: no faces, no hands, and background-art language both models understand deeply.

Our most-copied anime prompt — an empty street, one human trace, an era anchor doing the work.

Why it works: ”90s anime cel animation style” plus “painted background quality” anchors the era, and the character-free scene — just a schoolbag — skips the anime-face lottery.

The moody vertical companion piece, built for phone wallpapers.

Why it works: “seen from behind” sells the lonely mood and hides the face, while “neon signs reflecting in puddles” forces the wet-street glow.

The same painted-background idea on SDXL, tuned for brushwork.

Why it works: “hand-painted anime film background” plus “gouache texture” steers SDXL away from digital cel sharpness — and keep CFG at 6; 7.5 hardens the brushwork right back.

Summer festival street (Midjourney):

summer night festival street lined with paper lanterns and food stalls, yukata-clad visitors blurred in the midground, fireworks blooming over the rooftops, hand-painted background art, warm lantern gold against a deep indigo sky --ar 16:9 --stylize 250 --niji 6

Keeping the crowd “blurred in the midground” gives you festival energy without a dozen tiny broken faces.

Rural train crossing (Midjourney):

rural railway crossing at golden hour, a single train car passing, rice fields stretching to hazy blue mountains, telephone wires crossing an amber sky, nostalgic 90s cel anime style, painted clouds, subtle film grain --ar 16:9 --stylize 200 --no people --niji 6

--no people makes this a pure background plate — the shot a studio holds for three quiet seconds.

School rooftop at sunset (Midjourney):

empty school rooftop at sunset, chain-link fence silhouetted against a blazing orange sky, distant city haze, a forgotten canned drink on the concrete, melancholic anime key visual, long shadows --ar 16:9 --stylize 300 --niji 6

“Anime key visual” pushes poster-grade composition; the forgotten drink can is the human trace.

Snowy shrine steps (SDXL):

stone steps climbing to a small hilltop shrine in falling snow, red torii gate, paper lanterns glowing warm against blue winter dusk, anime film background art, soft painted snow texture, quiet lonely atmosphere
Negative: photorealistic, photograph, 3d render, people, text, watermark, blurry, oversaturated

32 steps · CFG 6 · DPM++ 2M Karras · 2:3 — the warm lanterns against blue dusk are the entire mood; keep both halves of the contrast.

Seaside hill town (SDXL):

steep seaside town street in midsummer, white walls and blue rooftops descending to a glittering harbor, towering cumulus clouds, hand-painted anime background art, saturated summer greens and blues, empty street
Negative: photorealistic, photograph, 3d render, people, cars, text, watermark, blurry

30 steps · CFG 6 · DPM++ 2M Karras · 16:9 — “empty street” plus “people, cars” in the negatives is what keeps the frame still.

Anime character and portrait prompts

Characters are where anime prompting gets hard. The pattern across all seven: identity from props and palette, and faces and hands managed deliberately.

Start here on SDXL — the era anchor and the negative-prompt guard in one prompt.

Why it works: “modern seinen manga style” pulls mature, detailed linework instead of the default big-eye look, while “western cartoon” and “chibi” in the negatives guard the other flank.

For characters you plan to reuse, generate a turnaround sheet first.

Why it works: “character reference sheet” with “front and back views” carries the layout — two views stay consistent where four collapse — and her identity lives in three props you can swap to redesign her.

Street musician (Midjourney):

full body anime illustration of a street musician girl with a worn acoustic guitar, oversized cardigan, headphones around her neck, singing under a station underpass light at night, modern anime film still, soft cel shading, clean lineart --ar 2:3 --stylize 150 --niji 6

The “cel shading, clean lineart” pair is what stops Niji drifting into western semi-realism on character shots.

Kneeling knight (Midjourney):

anime knight in ornate silver armor kneeling in a field of white wildflowers, long cape pooling on the grass, overcast soft light, solemn fantasy anime key art, detailed armor with painted reflections --ar 2:3 --stylize 200 --no photorealism --niji 6

A kneeling pose with a pooling cape hides hands and feet — the two failure points of full-body characters.

Elderly tea master (SDXL):

anime portrait of an elderly tea master in a sunlit wooden workshop, kind weathered face, steam rising from a clay teapot, detailed ink-and-watercolor shading, muted greens and warm wood tones, gentle expression
Negative: photorealistic, 3d render, western cartoon, chibi, extra fingers, deformed hands, blurry, text, watermark

30 steps · CFG 7 · DPM++ 2M Karras · 2:3 — aging the subject dodges the same-face problem that haunts young anime portraits.

Starship pilot (SDXL):

anime bust portrait of a starship pilot in an orange flight suit, scuffed helmet under one arm, hangar lights bokeh behind, confident grin, 90s OVA cel style with retro film grain, hard shadows, limited color palette
Negative: photorealistic, 3d render, western cartoon, chibi, extra fingers, deformed hands, blurry, text, watermark, oversaturated

30 steps · CFG 7 · DPM++ 2M Karras · 2:3 — “helmet under one arm” gives the visible hand a job; occupied hands render far cleaner than free ones.

Forest archer (SDXL):

full body anime illustration of a forest ranger drawing a longbow, hooded moss-green cloak, dappled light through the canopy, dynamic three-quarter stance, clean modern anime linework, flat cel shading
Negative: photorealistic, 3d render, western cartoon, chibi, extra fingers, deformed hands, extra limbs, blurry, text, watermark

32 steps · CFG 7 · DPM++ 2M Karras · 2:3 — a drawn bow is the genre’s hardest hand pose; keep “extra limbs” in the negatives and expect rerolls.

Chibi and kawaii mascot prompts

Chibi is its own dialect: thick outlines, flat pastel color, and proportions fragile enough that a stylize value can break them.

The template every sticker sheet on this site is built from.

Why it works: naming all six animals stops Niji inventing hybrid creatures to fill the slots, and stylize holds at 100 — higher, and chibi proportions drift toward full anime characters.

Chibi food characters (Midjourney):

sticker sheet of five chibi food characters with tiny arms and happy faces, a rice ball, a strawberry, a slice of buttered toast, a dumpling, and a pudding cup, thick clean outlines, flat pastel colors, white background, kawaii sticker style --ar 1:1 --stylize 100 --niji 6

Same trick, food edition — every item is named so nothing gets improvised.

Chibi weather spirits (Midjourney):

set of four chibi weather spirits, a beaming sun, a teary rain cloud, a sleepy crescent moon, and a mischievous lightning bolt, simple rounded shapes, thick outlines, soft pastel palette, white background, kawaii sticker style --ar 1:1 --stylize 100 --no gradients, 3d --niji 6

The --no gradients, 3d chain keeps the shapes print-flat for die-cut stickers.

Chibi baby dragon (SDXL):

single chibi baby dragon mascot, round teal body, tiny wings, big sparkling eyes, hugging a gold coin, thick clean outline, flat cel colors, plain white background, kawaii mascot design
Negative: photorealistic, 3d render, realistic scales, extra limbs, deformed, complex background, text, watermark

28 steps · CFG 7 · DPM++ 2M Karras · 1:1 — “complex background” in the negatives does more for sticker use than anything in the positive prompt.

Anime food and cozy interior prompts

Food and interiors are anime’s comfort genres, and both run on two tricks: glossy painted highlights and two-temperature lighting.

The food prompt everything else in this section is calibrated against.

Why it works: “painted highlights on the broth” makes the bowl look edible — cut it and the soup goes matte; if the chopsticks mangle an implied hand, switch to an overhead flat lay.

Proof that interiors carry mood through light alone.

Why it works: dusk lives in the two-light contrast — warm desk lamp against deep blue window light. Drop either phrase and Niji renders flat daytime.

Bakery at opening time (Midjourney):

small anime bakery interior at opening time, golden morning light through the shopfront glass, trays of glossy melon buns and croissants, handwritten chalkboard menu, steam curling from a coffee machine, slice-of-life anime background art, warm inviting palette --ar 16:9 --stylize 200 --niji 6

Glossy highlights do the appetite work here exactly as they do on the ramen broth.

Café window seat in the rain (Midjourney):

window seat of a quiet cafe on a rainy afternoon, raindrops streaking the glass, a steaming latte and an open notebook on the table, blurred umbrellas passing outside, modern anime film still, muted blues with warm interior light --ar 16:9 --stylize 250 --niji 6

Warm interior against cold rain is the bedroom’s two-temperature trick, relocated to a table for one.

Kotatsu winter living room (SDXL):

cozy Japanese living room in winter, a kotatsu heated table with a bowl of mandarin oranges on top, snow falling past the window, warm lamp glow, a cat curled on a floor cushion, soft cel shading, slice-of-life anime background art
Negative: photorealistic, photograph, 3d render, people, western cartoon, text, watermark, blurry

30 steps · CFG 6.5 · DPM++ 2M Karras · 16:9 — the mandarin oranges are the specificity that sells the season; generic “cozy room” prompts have no such anchor.

Overhead bento box (SDXL):

overhead anime illustration of an open bento box, white rice with a pickled plum, rolled omelet slices, octopus-cut sausages, broccoli and cherry tomatoes, glossy painted highlights, warm wooden table, anime film food art
Negative: photorealistic, photograph, 3d render, hands, chopsticks, text, watermark, blurry, oversaturated

30 steps · CFG 6.5 · DPM++ 2M Karras · 1:1 — overhead framing plus “hands, chopsticks” in the negatives sidesteps finger anatomy entirely.

Mecha and action prompts

The hardest section: action means bodies in motion and machines full of detail. Staging — silhouettes, occupied hands, human scale references — matters more than style words here.

The scale lesson every giant-robot prompt should copy.

Why it works: scale is carried entirely by “tiny engineers on catwalks dwarfed by its torso” — remove the people and the mecha reads as a desk model. The 40 steps matter; panel detail mushes at 28.

Mecha in a flooded city (Midjourney):

giant mecha wading through a flooded city avenue at dawn, water cascading off its armor plates, morning mist between drowned buildings, birds scattering from its shoulders, 90s mecha anime production art, muted steel blues with one red beacon --ar 21:9 --stylize 250 --niji 6

“Water cascading off its armor plates” communicates weight better than any size adjective.

Rooftop courier sprint (Midjourney):

anime action still of a courier sprinting across nighttime rooftops, dynamic foreshortened pose, motion blur on the background, speed lines, billboard glow rim-lighting the figure, bold 2000s action anime cel style --ar 16:9 --stylize 300 --niji 6

Motion blur goes on the background, not the figure — the character stays crisp while the world sells the speed.

Sunset sword duel (SDXL):

two anime swordfighters mid-duel on a temple rooftop at sunset, blades crossing in silhouette against a huge orange sun, roof tiles scattering, dramatic low angle, high-contrast anime key visual, speed lines in the clouds
Negative: photorealistic, 3d render, western cartoon, extra limbs, extra fingers, deformed hands, blurry, text, watermark

32 steps · CFG 7 · DPM++ 2M Karras · 16:9 — silhouette staging hides faces and hands at once, the highest-success action framing on SDXL.

Retro space battle (SDXL):

retro anime space battle wide shot, a battered cruiser trailing vapor past a ringed planet, escort fighters in formation, hand-painted starfield, 80s OVA cel style with film grain, dramatic rim light from a distant sun
Negative: photorealistic, 3d render, cgi, lens flare, text, watermark, blurry, oversaturated

32 steps · CFG 6.5 · DPM++ 2M Karras · 21:9 — “cgi” in the negatives is load-bearing; space scenes default hard to modern 3D rendering.

Battle mage (SDXL):

anime battle mage unleashing a spiral of blue flame, robes whipping in the shockwave, glowing sigils circling her raised staff, dramatic upward camera angle, crisp modern anime cel shading, high contrast lighting
Negative: photorealistic, 3d render, western cartoon, extra fingers, deformed hands, extra limbs, blurry, text, watermark

30 steps · CFG 7 · DPM++ 2M Karras · 2:3 — the raised staff gives both hands a job, and the upward angle hides where faces most often go wrong.

How to adapt these prompts and keep a character consistent

Lock a character with a descriptor block. Write three or four concrete identifiers — “silver bob haircut, round amber glasses, mustard-yellow scarf, navy duffel coat” — and paste the block verbatim into every prompt, changing only scene and pose. Vague traits re-roll every generation; concrete props survive. On Stable Diffusion, also fix the seed once you like a result — the consistency tool Midjourney lacks — and for long-running characters, generate a turnaround with the witch prompt above as ground truth.

Know each model’s failure mode. Midjourney’s is semi-realism creep — character shots slowly gaining rendered skin and western proportions; append “flat cel shading, clean lineart” and lower --stylize toward 150. SDXL’s is hands: the negative baseline handles most of it, staging the rest — occupy hands with props, hide them in poses, or crop above the wrists.

Which AI model is best for anime art?

Midjourney’s Niji mode is the best choice for polished anime images with minimal effort; Stable Diffusion XL is better when you need control. Niji delivers gallery-ready scenery and key visuals from short prose prompts. SDXL asks for more setup but gives you fixed seeds for consistent characters, reliable reference sheets, and anime-tuned community checkpoints. People who make anime art seriously end up using both: Niji to explore, SDXL to iterate.

Every prompt above ships with settings and an example image — browse the Midjourney prompt library and the Stable Diffusion library for more, or stay in genre at the anime style hub.