When Midjourney first launched, they started with just 500 users, and each of those people could invite one more, so it quickly grew to 1,000 users total. David Holz, the founder, encouraged everyone to share their creations on social media, which really helped get the word out and build a buzz.

They released the upscaling and variation buttons along with a new model, which was a big deal. With feedback from the community, the Midjourney team finally nailed down a solid pricing plan—before that, people were just generating images for free. That’s when they switched to a paid beta, and after that, the waitlist really started to grow.

Midjourney V3 dropped, and with it came the --stylize and --quality parameters, which made it way easier to fine-tune your images. They also released a new upscaler that seriously cut down on distortion and artifacts. Oh, and get this—the Midjourney Discord blew up to over a million users, even passing Fortnite and Minecraft’s servers. Pretty wild, right?

MidJourney V4 really took things to the next level. This version knows more about creatures, places, and details, and it’s much better at nailing the little things, no matter how complex the prompt. It handles multi-object and multi-character scenes like a pro and supports features like image prompting and multi-prompts. They added the --chaos parameter, where you can set it from 0 to 100 to control how wild the image grids get.

Midjourney dropped their anime model, Niji, and people were loving it. They built it with the anime studio SpellBrush, and it was miles ahead of anything else out there at the time. David Holz said the new Niji model was fine-tuned specifically to nail anime and illustrative styles—and it really shows!

Midjourney V5 really stepped things up again being more “unopinionated” compared to V3 and V4. It was the second model trained on their AI supercluster, and they spent 5 months developing it. V5 brought in much sharper image quality, more detailed visuals, and handled image prompting better. It supported --tile for seamless tiling, --ar for aspect ratios over 2:1, and --iw to control how much weight image prompts have compared to text prompts. A standout upgrade!

Midjourney rolled out the next version of Niji, and it set the standard for having a Niji version with every base model release. This one was really dialed into anime aesthetics, and the style was even stronger than the previous version. You could tweak it using the --stylize command to get the look just right. Then, over the next few weeks, they dropped set style options like 'expressive,' 'cute,' and 'scenic' to give you more control over the vibe.

Just a few months after V5, the team dropped Version 5.1. This one was a bit more opinionated, kind of like V4, and it worked way better with short prompts—super user-friendly. They also added something called "RAW Mode," which is basically an unopinionated setting like the default V5, so you could still get that flexibility. The big upgrades in 5.1 were higher coherence, better accuracy with text prompts, fewer weird borders or text glitches, and overall sharper images. It’s a pretty solid update!

Right after V5.1, they dropped Version 5.2, which came with some sweet new features. They introduced a whole new Aesthetic System that gave you even sharper images, better overall looks, and improved how well it understood text. Plus, it bumped up the diversity in the outputs. A "High Variation Mode" was added which makes variations way more different from each other. They also added a /shorten command so you can figure out which words in your prompt matter most, and finally, the long-awaited "Zoom Out" feature that lets you reframe your images. Definitely a fun upgrade!

After 9 months in the works, Version 6 was released, and it’s a game-changer. It’s the third model built on the AI Supercluster, and the image quality is next-level. What’s really cool is that it works even better with simpler prompts now, and the way it understands prompts is way sharper. David even said, “You’ll have to relearn how to prompt,” and to avoid the usual stuff like “award-winning, photorealistic, 4K, 8K”—basically, all that fluff. Upscales are twice as fast now, text rendering got a boost, and it performs way better with high --stylize values. It’s seriously impressive!

Sticking with the tradition of dropping a new Niji version after each base model, they released Niji Version 6 too. This one’s got an even stronger anime style than the earlier versions. If you want something more subtle, though, they made the --style raw option available, which helps tone things down a bit.

They dropped another update, Version 6.1, and it came with a bunch of cool improvements. The images are way more coherent, the overall quality is a big step up, and it handles small details like eyes, tiny faces, and hands way better. The new upscalers really improved the image and texture quality too, plus the text accuracy got a boost. They also added a personalization model that brings more nuance, surprise, and accuracy to the results. David even said, "Things should look generally more beautiful," and honestly, he’s right!