Niji 7 Prompts: 30 Copy-Paste Anime Prompts (Niji vs Midjourney, 2026)

TL;DR: A Niji 7 prompt is a literal description plus --niji 7. Write concrete nouns for the subject, the setting and the drawing style, then set --ar and --stylize. Niji 7 launched January 9, 2026 and is deliberately more literal than Niji 6, so vague mood prompts underperform. --sref works. --cref does not.

What are the best Niji 7 prompts, and how do you write one?

The best Niji 7 prompts are specific about three things in order: who or what is in the frame, where they are, and how the drawing is made. Then you append --niji 7. Midjourney's own Niji 7 launch post gives the instruction directly: "To make sure you are using Niji 7, type --niji 7 after your prompt in discord," or pick Niji 7 from the Version dropdown on the web.

Here is the shape, and a working example you can paste right now.

a girl with short lime green hair, hair in a single bun, blue eyes, wearing a
school uniform at a train station, anime screenshot
--niji 7 --ar 16:9 --s 150

That prompt is not ours. It is one of the example prompts Spellbrush published in the Niji 7 launch post, and it is worth studying because of what it does not contain. No "masterpiece". No "8k". No "trending on artstation". Just a described character, a described place, and one style anchor.

This guide covers 30 prompts across distinct anime and illustration styles, the parameters --niji 7 actually accepts according to Midjourney's documentation, and the honest cases where you should point the prompt at the default Midjourney model instead. Every parameter claim here is sourced to a Midjourney-owned page with the date we checked it.

What changed in Niji 7, and why did your old Niji prompts stop working?

Niji 7 became more literal. That single change breaks more inherited prompts than anything else in the release.

Midjourney's Version article describes Niji 7, launched January 9, 2026, as a "major boost in coherency" where "fine details like eyes, reflections, and small background elements are now much clearer," and then adds the warning: "This version is more literal, so broad or 'vibey' prompts may not behave the same as before." Spellbrush put it more bluntly in the launch post: "Since the new model gives you more of exactly what you want, you'll find that your vibes-y prompts may not work as well as they used to."

They illustrated it with a genuinely funny failure. The prompt anime screencap, a remote spring was meant to produce a secluded hot spring. Niji 7 read "remote" as a remote control and drew one, in springtime.

The second change is aesthetic. Per the launch post, Spellbrush "toned down the amount of rendering in the default style (you can think of it as 3D-ness)" to expose the line work underneath. Midjourney's Version article calls it "a cleaner, flatter look designed to highlight its improved line work." If your Niji 6 prompts leaned on heavy rendering and glossy shading, Niji 7 will hand you something flatter unless you ask for the rendering explicitly.

What is the Niji 7 prompt skeleton?

Use five slots, in this order. Everything below in this guide is built from it.

1. Subject — age or archetype, hair colour and cut, eye colour, one distinguishing feature.
2. Wardrobe and props — named garments, not "cool outfit".
3. Setting and time — a place and a light condition.
4. Drawing style — the medium words that tell Niji how it is drawn.
5. Parameters — --niji 7, then --ar, then --s.

The style slot is doing more work than people expect. Spellbrush names two style triggers in the launch post itself: use the phrase "anime screenshot" to see Niji 7's line work, and "minimalist graphic logo" to see its handling of flat space. Both behave like style presets you type in plain English.

On --stylize: Midjourney's Stylize article puts the default at 100 with a range of 0 to 1000, describing low values as sticking to your prompt and high values as giving the model more creative freedom. For a literal model like Niji 7, that maps cleanly. Character sheets and logos want 0 to 150. Atmospheric key art wants 300 to 600.

30 copy-paste Niji 7 prompts by style

Replace the bracketed parts. Each one is a starting point, not a finished artwork.

Anime screencap and TV-still looks

a teenage boy with messy black hair and amber eyes, red track jacket, standing
on an empty school rooftop at late afternoon, long shadows, anime screenshot
--niji 7 --ar 16:9 --s 150

two students sharing an umbrella on a wet crosswalk, city buses behind them,
overcast grey light, puddle reflections, anime screenshot, film grain
--niji 7 --ar 16:9 --s 200

a close-up of a girl's face turning toward the camera, wind in her hair,
sunset backlight, single tear, dramatic anime screenshot, high contrast
--niji 7 --ar 16:9 --s 250

Shoujo and soft portrait

a young woman with long lavender hair and grey-green eyes, white blouse with a
navy ribbon, seated by a window with sheer curtains, soft morning light,
shoujo manga illustration, delicate line work, pastel palette
--niji 7 --ar 2:3 --s 250

portrait of a girl holding a bouquet of white anemones, freckles, gentle half
smile, warm neutral background, soft coloured pencil texture, shoujo style
--niji 7 --ar 4:5 --s 200

Shounen action

a swordsman mid-leap with a chipped katana, torn haori, sparks trailing behind
the blade, dust cloud below, dynamic low angle, shounen manga action panel,
speed lines
--niji 7 --ar 3:4 --s 300

a martial artist bracing against a shockwave, cracked stone under both feet,
clenched fists, hard rim light, shounen anime key visual, bold ink outlines
--niji 7 --ar 16:9 --s 350

Manga panel and screentone

black and white manga panel, a detective standing under a flickering streetlamp,
trench coat, cigarette smoke, heavy screentone shading, crosshatching,
no colour, ink on paper
--niji 7 --ar 3:4 --s 100

four-panel black and white manga page, a cat repeatedly knocking a mug off a
desk, comedic reaction faces, screentone backgrounds, gutter spacing between
panels
--niji 7 --ar 2:3 --s 120

Chibi and kawaii

chibi character, small round body, oversized head, mint green hair in twin
buns, star-shaped hairpin, holding a strawberry milk carton, plain cream
background, thick clean outlines, sticker style
--niji 7 --ar 1:1 --s 180

a sheet of nine chibi expression icons of the same character, happy, sleepy,
angry, crying, smug, surprised, flat colours, white background, even grid
spacing
--niji 7 --ar 1:1 --s 100

Retro cel animation

1990s cel animation still, a woman in a red blazer answering a corded telephone
in a cluttered office, warm CRT colour cast, visible cel paint texture, slight
registration offset
--niji 7 --ar 4:3 --s 250

retro anime OVA frame, a fighter pilot in an open cockpit at dusk, orange sky,
grainy film scan, muted 1980s palette, hand-painted background
--niji 7 --ar 16:9 --s 300

Watercolour and storybook illustration

children's book illustration, a small girl in yellow rain boots leading a
sheepdog across a stone bridge, loose watercolour washes, visible paper grain,
soft edges, muted greens
--niji 7 --ar 4:5 --s 300

storybook spread, a fox family sharing tea inside a hollow tree, warm lantern
light, gouache texture, hand-lettered feel, generous negative space
--niji 7 --ar 16:9 --s 280

Painted anime backgrounds

detailed anime background art, a rural train platform at golden hour, rice
fields on both sides, cicada haze, telephone poles receding into the distance,
no characters
--niji 7 --ar 16:9 --s 350

detailed anime background art, a cramped second-hand bookshop interior, dust in
the light shaft, stacked paperbacks, ceiling fan, no characters
--niji 7 --ar 3:2 --s 300

Cyberpunk and neon

a courier in a translucent raincoat on a neon-lit night street, holographic
shop signs in mixed scripts, wet asphalt reflections, anime cyberpunk
illustration, cool blue and magenta palette
--niji 7 --ar 16:9 --s 400

a hacker's cramped apartment at 3am, six monitors casting blue light, cables
across the floor, instant noodles going cold, anime illustration, high
contrast, deep shadows
--niji 7 --ar 16:9 --s 300

Mecha and hard surface

a humanoid mecha kneeling in a flooded hangar, panel lines and rivets, coolant
steam, technician on a scaffold for scale, mechanical anime illustration,
industrial palette
--niji 7 --ar 16:9 --s 300

Character sheet and turnaround

character reference sheet, teenage archer, short black hair, scar over the left
eye, dark green cloak, front view, three-quarter view and back view, neutral
pose, clean line art, flat white background
--niji 7 --ar 16:9 --s 50

costume design sheet for a shrine maiden, three outfit variations side by side,
labelled fabric swatches, flat colour, white background, clean line work
--niji 7 --ar 16:9 --s 80

Flat graphic and logo work

minimalist graphic logo, a stylised crane folded from paper, two colours,
geometric, thick even strokes, flat white background
--niji 7 --ar 1:1 --s 100

flat vector anime poster, a girl on a bicycle against a single large sun shape,
three-colour palette, no gradients, bold shapes, large areas of empty space
--niji 7 --ar 2:3 --s 150

Ink, brush and traditional media

sumi-e ink brush illustration in an anime style, a lone swordsman on a
mountain path, wet brush strokes, spare composition, off-white paper, single
vermilion accent
--niji 7 --ar 2:3 --s 300

ukiyo-e influenced modern anime illustration, a woman under a paper umbrella in
falling snow, flat colour blocks, visible woodblock registration lines, muted
indigo and cream
--niji 7 --ar 2:3 --s 350

Text on the image

Midjourney's Text Generation article is explicit: put the words inside double quotation marks, single quotes will not work, and if the text comes out wrong, lower the stylize value.

light novel cover illustration, a girl in a school uniform holding a glowing
sword, night sky behind her, with the words "SUMMER OF STEEL" written across
the top in bold letters
--niji 7 --ar 2:3 --s 120

anime style café menu board, chalk lettering with the words "TODAY'S SPECIAL"
in large text, small pastry drawings around the border, warm wood background
--niji 7 --ar 3:4 --s 100

Vertical and webtoon formats

vertical webtoon panel, a girl scrolling her phone on a night bus, interior
lights, sleeping passengers, soft digital colouring, clean outlines, room for
dialogue at the top
--niji 7 --ar 9:16 --s 200

phone wallpaper illustration, a cat curled on a windowsill above a rainy
street, warm interior light against cool exterior, flat anime style, uncluttered
composition
--niji 7 --ar 9:16 --s 250

Which parameters actually work with --niji 7?

This is where most Niji guides go stale, so here is the documented state as of August 26, 2026. Sources are Midjourney's own help centre articles, each linked.

Three of those deserve a sentence each.

The --iw ceiling is genuinely different. Midjourney's Image Prompts article publishes a three-column table: image weight defaults to 1 on Version 8.1, Version 7 and Niji 7, but the range is 0 to 3 on the first two and 0 to 2 on Niji 7. A --iw 2.5 that behaves on V8.2 is out of range the moment you switch the flag.

--raw on Niji is community lore, not documentation. Midjourney's Raw article says only that "Raw is compatible with Midjourney versions 5.1 and later" and never names a Niji version. The Version compatibility chart has no Niji column at all. The Niji 7 launch post says most features were carried over but does not list Raw among them. Plenty of users report that it works and produces flatter, less decorated output. We could not verify it from a Midjourney-owned page, so treat it as reported rather than documented and test it on your own prompt. Our separate write-up on whether --style raw still works covers the wider syntax change, since --style raw itself is now legacy syntax and --raw is the current parameter.

There is no Niji 8. Midjourney's compatibility chart pairs --niji 6 with V6 and --niji 7 with V7, and the V8.1 and V8.2 column has no Niji companion. If a post tells you to type --niji 8, it is guessing. The full flag list lives in our Midjourney version flags reference.

Niji 7 or Midjourney V8.2: which should the prompt go to?

Point it at Niji 7 when the output should read as drawn. Point it at the default Midjourney model when it should read as photographed, or when you need a style outside anime.

The default model has moved on since this guide first ran. Per Midjourney's Version article, V8.2 has been the default since July 24, 2026. V8.1 shipped April 14, 2026 and held the default slot from June 10 to July 23. V7, the generation --niji 7 sits alongside in the compatibility chart, was the default from June 17, 2025 to June 9, 2026. V8.0 was an alpha and no longer renders.

Two practical notes on that table.

HD is a real gap, not a rounding error. Midjourney's Image Size & Resolution article states that at the default 1:1 ratio, V8.2 produces HD images at 2048 by 2048 pixels and SD images at 1024 by 1024. HD is not square-only, though: the Version chart caps HD at a 4:1 aspect ratio while SD goes to 14:1, which only makes sense if HD renders at many shapes. HD costs 1.3 minutes of GPU time against SD's 0.8. If you need print-scale output from an anime prompt, you will be upscaling Niji output rather than generating it at 2K.

The HD default has flipped twice, so do not assert one. Midjourney's V8.1 Alpha post on April 14, 2026 said HD was "so cheap we're making it default for V8.1." The V8.1 Updates post on April 30 reversed it: "To save compute during a server transition, we're making SD default for a bit." The most recent changelog, dated August 20, 2026, discusses HD selection bugs without restating a global default. Check your settings panel rather than trusting any article, including this one.

For photoreal prompts and the V8-generation parameter set, our Midjourney prompt examples guide and the aspect ratio reference cover the other half of the workflow.

How do you keep the same character across Niji 7 prompts?

Without --cref, consistency in Niji 7 comes from two things: a locked style reference and an unusually disciplined character description.

Lock the style first. --sref is the feature Spellbrush singled out in the launch post, and Midjourney's Style Reference article documents --sw from 0 to 1000 with a default of 100. Generate one image you like, then reuse it as the style reference for every subsequent frame.

[new scene description], anime screenshot
--niji 7 --sref [your-locked-image-url] --sw 150 --ar 16:9 --s 200

Then write the character as a fixed block of text and paste it unchanged into every prompt. Niji 7 being literal is an advantage here.

a girl, 17, short lime green hair in a single bun, blue eyes, one visible fang,
grey school blazer with a red tie, small crescent earring in the left ear,
[new setting], anime screenshot
--niji 7 --sref [your-locked-image-url] --sw 150 --ar 16:9 --s 200

Spellbrush signalled a replacement is coming: "we have a super special secret surprise in the works that we think you will love more than the old cref." As of August 26, 2026 that feature has not been documented, so plan around the tools that exist. The same launch post listed Personalization and moodboards for Niji 7 as "coming soon" in January 2026, and Midjourney's Personalization article still scopes --p to "Midjourney versions 6 and later" without naming Niji 7, so we would not promise readers it works there either.

If you need a face locked with reference-image precision rather than description, that is a job for the default model and --oref, which is documented as Midjourney V7 only and costs double GPU time. Our guide to character consistency with --cref and --oref walks through both paths.

What Niji 7 prompt mistakes cost people the most time?

1. Pasting a Niji 6 prompt unchanged. The vibes-based phrasing that carried Niji 6 gets read literally now. Rewrite the mood into objects and light.
2. Typing --cref and assuming it silently worked. It is not supported. You will get a plausible but unrelated character and blame the prompt.
3. Reaching for --style cute. Niji 5 syntax. Describe the cuteness instead: rounded proportions, large eyes, pastel palette, thick outlines.
4. Expecting Raw to produce a photograph. Niji stylizes by definition. Even if Raw behaves on Niji, it reduces decoration, it does not add realism.
5. Setting --iw 3. Out of range on Niji 7, which stops at 2.
6. Running one stylize value for every job. Character sheets and logos want it low. Atmospheric key art wants it high. Test both ends of the same prompt.
7. Fighting Niji for photoreal output. There is no flag for it. Switch models. Our post on why Midjourney prompts fail covers the wider version of this mistake.
8. Using single quotes for on-image text. Midjourney's Text Generation article says double quotes only, and to lower stylize when letters garble. More on that in why text in AI images comes out garbled.

How do you build a repeatable Niji 7 workflow?

Stop rewriting parameters from memory. Build a small set of prompt templates with the flags already attached, then vary only the description.

1. Pick four styles you actually ship. Not twelve. For most people it is portrait, background, character sheet and one graphic format.
2. Fix the parameter tail for each. Character sheet gets --niji 7 --ar 16:9 --s 50. Key art gets --niji 7 --ar 16:9 --s 350. Write them once.
3. Save one locked --sref per project, not per image.
4. Run the same description on both models once, --niji 7 against the current default, and score the result. Do it once per new shot type, then stop re-litigating it every session.
5. Store the winners somewhere searchable. A prompt you cannot find in thirty seconds is a prompt you will rewrite from scratch.

That last step is where most anime workflows leak time. Prompt Architects does not generate images. It structures and stores the prompt you paste into Midjourney, so the parameter tail, the --sref code and the fixed character block travel with the template instead of living in your scrollback. If you would rather see the style-word half of the problem, our Midjourney style modifiers reference catalogues the vocabulary.

The short version

Niji 7 is the current anime model, launched January 9, 2026, and it is more literal than the model it replaced. Write concrete nouns. Use --sref for consistency because --cref is gone. Keep --iw at or below 2. Do not type --niji 8, and do not expect a photograph. When the brief calls for photoreal rendering or 2K HD output, switch to the default Midjourney model, currently V8.2, and let Niji do the thing it is genuinely best at.