I have published more than 3,700 AI artworks since 2023 — and like every creator, I could never count the real number behind that: the prompts, the dead ends, the try-again-at-2am images that nobody will ever see. A good part of that journey happened inside Midjourney, a tool I like and use a lot. This article is about where it stands in 2026 — what changed, what improved, and one thing I honestly feel it lost.
For newcomers, the short version: Midjourney is an AI image generator. You describe what you imagine in words — a prompt — and the model paints it. Since the beginning it has been known for images with a strong sense of light, composition and mood, and that identity is still recognizable today.
This year brought the biggest technical jump in a while: the V8 era. In March 2026, Midjourney launched V8 in alpha, rebuilt from the ground up with a new dataset and new training infrastructure. Generations became roughly five times faster, and a new HD mode brought native 2K resolution. V8.1 followed in April and became the default model in June, replacing V7 — with HD mode about three times faster and cheaper, standard generations 50% faster and 25% cheaper, and much better text rendering, historically one of the tool's weakest spots.
Speed sounds like a boring metric until you feel it. An image that used to take 30 to 60 seconds now appears in under ten. Iteration is the heart of AI art, and when the feedback loop collapses from a minute to seconds, you explore ideas you would never have bothered to try before.
The exploration tools grew too. Draft mode, introduced in June, generates 24 low-resolution images while consuming half the fast hours — useful for hunting compositions before committing to HD. Combined with the new --sref random parameter, which pulls 24 different style references at once, it turns style discovery into something closer to play. Midjourney offers video generation too — any image can become a short clip. It is good for many cases, but honestly, it is becoming obsolete: there have been no updates to it in a long time, while the dedicated video tools keep moving fast. I cover that world in a separate article on this blog.
Something that would have sounded impossible in the old days: Midjourney models are no longer exclusive to the Midjourney platform. You can now find them inside other services too. I honestly have not tested those versions enough to say whether the results differ from the original platform — but the walls around the garden are clearly coming down, and that alone says a lot about how the market changed.
And here is the thing I need to say, because honesty is the point of this article. Midjourney keeps getting better and better in quality — sharper, faster, more controllable with every version. But in the bottom of my heart, I feel it lost something along the way: the creativity it used to have. In the early days, Midjourney surprised us constantly. You would write a simple prompt and receive something you never asked for and never imagined — and sometimes that accident was the artwork. Today the results are technically superior and far more obedient, but they surprise me less. Maybe that is the natural price of maturity. I still miss it.
What truly keeps me on the platform is niji — the anime-focused model family available inside Midjourney. I love niji: its sense of character, color and expression has a personality of its own, and for the kind of portraits I make it remains special. It deserves an article of its own, and I am already preparing one.
So, is Midjourney worth it in 2026? For me, yes — it is good, I like it, and I use it a lot. I will not tell you it is 'the best', because I do not believe that word means much for creative tools. Try it with draft mode, learn style references early, and treat prompts as conversations rather than commands. The machine got faster and sharper — bringing the surprise back is our job now.