If you make anime or illustration, you have hit this fork: do you use Midjourney’s general model or its Niji model? They live in the same app, take nearly the same prompts, and produce very different pictures. Picking wrong means fighting the model for a look it was never built to give. So let us settle Niji vs Midjourney with concrete examples and a clear rule for when to reach for each.

Short version up front: Niji is the anime and illustration specialist; the general Midjourney model is the all-rounder that leans photographic and cinematic. For anime, manga, character art and stylized illustration, Niji usually wins. For photoreal, concept art, product and design, the general model usually wins. The rest of this is the detail behind that rule.

Niji vs Midjourney: what they actually are #

Both are models inside the same Midjourney service. You choose between them with a single parameter at the end of your prompt. The general model is invoked with --v (for example --v 6.1). Niji is invoked with --niji (for example --niji 6). Same prompt box, same commands, different brain doing the drawing.

Niji was trained with a heavy emphasis on anime and illustration aesthetics. It understands the visual grammar of that world — cel shading, expressive eyes, dynamic poses, the conventions of manga and Japanese illustration — in a way the general model approximates but does not nail. The general model is broader. It is exceptional at photorealism, cinematic scenes, architecture and design, and it can do a competent anime look, but “competent” is the ceiling there, not “native.”

The same prompt, both models #

Nothing explains this faster than running one prompt through both. Take a simple character:

/imagine prompt: a young swordsman with silver hair, standing on a rooftop at sunset, wind in his coat, dramatic pose --ar 2:3 --v 6.1

On the general model you get something semi-realistic — plausible skin texture, photographic lighting, a slightly cinematic feel. It looks like a still from a live-action film or a realistic game. Now the same words on Niji:

/imagine prompt: a young swordsman with silver hair, standing on a rooftop at sunset, wind in his coat, dramatic pose --ar 2:3 --niji 6

Niji returns clean anime linework, flatter and more deliberate shading, larger expressive eyes, the proportions and energy of an anime key visual. Neither is better in the abstract. They are answers to different questions. If you wanted an anime character, Niji answered it and the general model missed; if you wanted a realistic one, the reverse.

Where Niji wins #

Reach for Niji when the target look lives anywhere in the anime and illustration family:

  • Anime characters and key visuals. Clean lines, cel shading, the genuine article rather than a photo trying to look drawn.
  • Manga panels. Black-and-white linework and screentone-style shading come far more naturally.
  • Stylized illustration. Light novel covers, visual-novel sprites, anime-style environments and backgrounds.
  • Expressive faces and emotion. Niji handles the exaggerated expression language of anime — wide eyes, blush, dynamic emotion — that the general model tends to dampen toward realism.
  • Cute and chibi work. Rounded, simplified, kawaii styles are squarely in Niji’s wheelhouse.
/imagine prompt: a cheerful witch girl flying on a broom over a pastel town, fluffy clouds, anime key visual, vibrant colors --ar 16:9 --niji 6 --style cute

If anime is your main lane, our dedicated anime AI art prompts guide goes deep on getting the most out of Niji specifically.

Where the general Midjourney model wins #

Reach for --v when realism, cinematic feel or design precision is the goal:

  • Photorealism. Portraits, product shots, food, anything meant to look like a photograph. This is the general model’s home turf.
  • Cinematic concept art. Film-still moods, dramatic realistic lighting, gritty sci-fi and fantasy environments with photographic texture.
  • Architecture and interiors. Real materials, accurate perspective, magazine-grade spaces.
  • Logos and graphic design. Flat marks, clean geometry and design work sit better on the general model.
  • Realistic environments. Landscapes and nature meant to look photographed rather than illustrated.
/imagine prompt: cinematic portrait of a weathered explorer in an icy cave, dramatic torchlight, photorealistic, 35mm --ar 16:9 --stylize 250 --v 6.1

For photoreal work specifically, our guide to photorealistic AI art prompts covers the lens and lighting language that gets the most out of the general model.

Niji’s style presets change everything #

One reason Niji is so flexible inside its lane is the --style parameter, which applies a named aesthetic preset. The same Niji prompt swings widely depending on the preset:

  • --style original — the baseline Niji look.
  • --style cute — softer, rounder, more charming and kawaii.
  • --style scenic — emphasis on backgrounds and atmospheric, cinematic environments.
  • --style raw — less automatic styling, more control handed back to your prompt.
/imagine prompt: a quiet seaside town at golden hour, anime background art, detailed --ar 16:9 --niji 6 --style scenic

That --style scenic preset is a quiet weapon for anime-style landscape and background art — it pushes Niji toward the lush, painted environments you see in anime films. Swap to --style cute and the same town turns soft and storybook. Experiment with presets before you conclude Niji “can’t” do a particular look; often it can, you just had the wrong preset on.

What stays the same across both #

Here is the good news: almost everything you know transfers. The prompt structure is identical — subject, descriptors, environment, lighting, style, parameters — regardless of which model you pick. The core parameters behave the same way too. --ar sets the frame on both. --stylize controls aesthetic strength on both. --chaos varies the grid on both. --style raw exists on both and does the same job of dialing back automatic styling.

So switching between Niji and the general model is genuinely just changing one flag at the end. You do not relearn prompting; you reuse it. If you want the full parameter reference that applies to both, keep our Midjourney parameters cheat sheet handy.

A practical decision rule #

When you are unsure which to use, ask one question: should this look drawn or should it look photographed?

  • Should look drawn, animated, or illustrated → Niji.
  • Should look photographed, filmed, or rendered realistically → general model with --v.

The edge cases are where it gets interesting. Semi-realistic illustration, painterly concept art, stylized-but-not-anime work — these can go either way, and the honest answer is to run both and compare. It costs you one extra job and removes all the guessing. I do exactly this whenever a piece sits in the gray zone: same prompt, one with --v 6.1, one with --niji 6, then pick the winner.

Can you blend the two looks? #

Not in a single job — a prompt runs on one model or the other, not both at once. But there are two practical ways to get a hybrid feel. First, use an image prompt: generate a base on one model, then feed that image into a new prompt on the other model with a tuned --iw to carry some of its character across. Second, lean on --style raw on Niji to pull it toward realism, or push --stylize down on the general model with explicit “anime style” language to nudge it toward illustration. Neither fully merges the two, but both let you meet in the middle when a project wants it.

So, which should you choose? #

For anime, manga, character illustration and anything that should look hand-drawn, Niji is the right tool and it is not close. For photorealism, cinematic concept art, design and realistic environments, the general Midjourney model wins just as decisively. The whole choice collapses to that single question of drawn versus photographed, and the switch costs you one parameter.

The best habit you can build is to stop treating it as a permanent allegiance. Niji and the general model are two tools on the same bench. Pick per project, run both when you are genuinely unsure, and you will get the right look far more often than someone locked into one model out of habit. To plan a full series of consistent characters across either model, read our guide to character portrait prompts, and use ArtPrompts Generator to draft prompts for both Niji and the general model side by side.