niji-journey is a separate series of AI image models built inside Midjourney, developed as a collaboration between Midjourney and Spellbrush, and aimed specifically at Eastern and anime aesthetics and illustrative styles. It isn't a filter or a style parameter bolted onto the main Midjourney model — Midjourney's own documentation describes it as its own series with a dedicated website and Discord server, run alongside the main Midjourney community rather than folded into it. There are also iOS and Android apps. If you've only used Midjourney's main model and typed --niji as an afterthought, it's worth knowing there's a whole parallel product here, with its own release cadence and its own team making calls about where to take the aesthetic.
The current version is Niji 7, which launched January 9, 2026. Per the official announcement, it had been a year and a half since the previous Niji version shipped, so this wasn't a minor point release — it's the first meaningful update to the anime model in a long stretch, and it changes enough about how the model behaves that prompts tuned for the older version won't necessarily transfer cleanly.
Activation is simple. In Discord, type --niji 7 after your prompt. On the web interface, select "Niji 7" from the Version dropdown. That's the whole process — there's no separate account or migration step described in either source.
Midjourney's own documentation is direct about where Niji 7 is better: it "brings a major boost in coherency," and fine details like eyes, reflections, and small background elements are "much clearer" than before. The niji-journey team frames this as their biggest win too, calling it "Crystal Clarity" and describing it plainly as an HD upgrade — their own examples cite reflected detail in eyes and individual floating flower petals rendering cleanly in dreamy backgrounds, the kind of small elements that tend to smear or blur in earlier anime-style generators.
The model also follows prompts more closely now. Midjourney's docs note this helps "with specific designs or repeatable characters," which is a meaningfully different value proposition than raw image quality — it's about being able to describe something precisely and have the model actually build that thing, rather than a loose interpretation of it. The niji-journey team calls this "Superior Prompt Following" and offers two of their own test prompts as examples: "a pair of cubes, the red cube is on the left and the blue cube is on the right," and a character "with four arms, each of them holding an ice cream." Both are the kind of compositional, spatially specific instructions that generative image models have historically struggled to execute literally, and both are presented as cases Niji 7 now handles.
The same trait that makes Niji 7 better at precise, repeatable designs makes it worse at the kind of loose, atmospheric prompting a lot of people actually use. Midjourney's documentation states this directly: "this version is more literal, so broad or vibey prompts may not behave the same as before." The niji-journey team's own announcement is even more candid about it, and they picked a genuinely embarrassing example to make the point: the prompt "anime screencap, a remote spring" produced a remote control sitting in a spring season, rather than a secluded body of water. Their conclusion, quoted directly: "your vibes-y prompts may not work as well as they used to."
That's worth taking at face value rather than glossing over. If your workflow leans on evocative, impressionistic prompts and lets the model interpret the mood, Niji 7 is more likely to take you literally and miss the vibe you were going for. If your workflow depends on specific, describable compositions — a particular character doing a particular thing, a defined layout — Niji 7 is the more capable version. Anyone switching from an older Niji version should expect to rewrite prompts that used to work, not just reuse them.
Niji 7 also introduces a cleaner, flatter default aesthetic, and the team is explicit that this was a deliberate design choice rather than an accidental side effect of other changes. Per their announcement, they pushed two tenets in this version: "The Beauty of Line" and "Simple is the Most Difficult." In practice, they toned down how much rendering — how much 3D-like shading and depth — appears in the default output, in favor of a flatter look intended to highlight the model's improved line work. To see it, they suggest prompting "anime screenshot" for the linework and "minimalist graphic logo" for the simplicity.
The reasoning they give for leaning into simplicity is worth quoting directly, because it's an unusually candid piece of model-design thinking from a vendor: "AI elements probabilistically fail, but simple drawings have very few elements. If the drawing has only 4 elements, and one is a failure, then it's a bad drawing... it is always easier to make a drawing with 1000 elements." They add that simple drawings require large areas of visual emptiness, which is inherently riskier for a generative model to render convincingly — and that they consciously tuned for simplicity anyway, accepting that risk. If you're expecting the heavily rendered, painterly look some earlier anime-model outputs leaned toward, the Niji 7 default is a deliberate step away from that.
Backwards compatibility is mostly intact. The niji-journey team says most features were ported forward into Niji 7, including the --sv 4 style value flag. Style Reference (sref) was specifically called out as one of their favorite ported features, suggesting it works as expected in the new version.
The notable exception is Character Reference (cref), which Niji 7 does not support. The team's own framing softens the blow without actually replacing the feature: they say they have "a super special secret surprise in the works that we think you will love more than the old cref." That's a promise about the future, not a description of anything shipped. As of the sources checked for this review, cref support in Niji 7 does not exist. If your process depends on generating a consistent character across many images, that's a real workflow gap right now, not a rumor to dismiss.
Two other features — Personalization and Moodboards — were described as "coming soon" at the time of the January 8, 2026 announcement, with a rating task said to precede them and personalization and moodboards following shortly after. Neither source used for this review confirms whether those features have since shipped, so treat that timeline as the team's stated intent at launch rather than a confirmed current state.
Here's a detail that matters more than it sounds like it should: niji support does not extend to Midjourney's newest main-model versions. Midjourney's own feature compatibility chart shows the Niji Version row supporting V6 (with --niji 6) and V7 (with --niji 7) — but V8.1 and V8.2 both show no niji version support at all. Midjourney's current default main-model version is V8.2, which became the default on July 24, 2026.
Put plainly: if you're using the current default Midjourney experience, there is no niji flag to pair it with. To use Niji 7, you either need to be running an older main-model version alongside it, or you use niji-journey's own dedicated website, Discord server, or mobile apps directly rather than going through the main Midjourney interface on its current default version. This is easy to miss if you assume niji is just a style parameter that rides along with whatever main version you're already on — per Midjourney's own documentation, it currently doesn't.
Niji 7 is a good fit if you specifically want anime and illustrative aesthetics with cleaner linework, better small-detail rendering, and prompts that describe a specific composition, character, or design you want reproduced accurately. It's also the right fit if you're comfortable working through niji-journey's own site, Discord, or apps rather than expecting it to show up automatically inside whatever main Midjourney version you use day to day.
It's a weaker fit if your process depends on cref for consistent characters across a project — that feature is simply gone for now. It's also a weaker fit if you're used to loose, mood-driven prompting and want the model to interpret intent rather than take you literally; Niji 7's own makers warn that this is exactly the kind of prompt that now underperforms. And it's not usable at all through the current default Midjourney experience on V8.2, since that main version doesn't support a niji flag.
Neither of the sources used for this review states current subscription pricing for Midjourney or niji-journey access, so this article won't estimate or repeat a number that isn't sourced. Check Midjourney's own site for current plan pricing before subscribing.
No. Per Midjourney's official feature compatibility chart, the Niji Version row shows V6 supporting --niji 6 and V7 supporting --niji 7, while V8.1 and V8.2 show no niji version support at all. Midjourney's current default main-model version is V8.2, which became the default on July 24, 2026. That means if you're on the current default Midjourney experience, you can't pair it with a niji flag — you'd need to run an older main version alongside niji, or use niji-journey's own dedicated site, Discord, or apps directly.
It's gone. The official niji-journey announcement explicitly states cref is not supported in Niji 7, while noting the team has "a super special secret surprise in the works that we think you will love more than the old cref." No replacement had shipped or been detailed as of the sources checked for this review. If cref-driven character consistency is central to your workflow, that's a real gap to plan around, not a rumor.
Because Niji 7 is deliberately more literal. Midjourney's Version documentation says Niji 7 "follows prompts more closely" and that "broad or vibey prompts may not behave the same as before." The niji-journey team's own announcement is blunter: it gives the example of the prompt "anime screencap, a remote spring" producing a remote control in a spring season rather than a secluded body of water, adding "your vibes-y prompts may not work as well as they used to." Loose, mood-board-style prompts that relied on the model filling in gaps are the ones most likely to disappoint; specific, literal prompts tend to do better.
niji-journey is a series of AI image models within Midjourney focused on Eastern and anime aesthetics and illustrative styles, developed as a collaboration between Midjourney and Spellbrush. It has its own dedicated website and Discord server separate from the main Midjourney community, plus iOS and Android apps. Niji 7, the current version, launched January 9, 2026.
In Discord, type --niji 7 after your prompt. On the web interface, select "Niji 7" from the Version dropdown. Most existing features carried over from prior Niji versions, including the --sv 4 style value flag and Style Reference (sref), which the niji-journey team called out as a favorite ported feature.
If you're weighing niji-journey against Midjourney's main model, our Midjourney review covers the flagship model directly, and our Midjourney V8 vs V7 breakdown details exactly which features the current default model dropped — including the Niji version switch. For the broader anime and illustration AI landscape, see our Best AI Image Generators roundup and our Best Free AI Tools guide if budget is the deciding factor. If you're building an AI-assisted content workflow around image generation, our AI Tools for Content Creators guide has more context, and our AI Tools Starter Guide is a good starting point if you're new to evaluating these tools at all.