There’s a reason why Midjourney has been the talk of the town in the AI world. Sometimes it’s because they’re once again roped into discussions about deepfakes. But today, it’s for something good: they finally released an update to their V6 model called, well, what else but Midjourney V6.1.
And trust me, it’s a game-changer.
So, what’s new with Midjourney V6.1? Well, the team has focused on improving the model’s creativity and making the images even more coherent and lifelike. I’ve been using it for weeks, and let me tell you, it might be a subtle change, but it makes a whole world of difference in the output quality.
Oh and yeah, they also added a pretty nifty feature that users have been begging for for the longest time.
But don’t just take my word for it. Let’s dive in and see how Midjourney V6.1 stacks up against its predecessor across different image categories.
What is Midjourney?
Let’s be honest — if you’re even remotely familiar with AI, then you’ve probably heard of Midjourney. This is an AI image generation model that can create any photo ranging from artwork to real (well, something that looks like one, at least) images. Released in 2021, Midjourney’s latest model is V6.1, which we’re reviewing today and comparing to V6.
For a more complete overview, here’s our review of Midjourney from a year back. It’s a great starter guide if you’re looking to familiarize yourself with the application.
What’s New With Midjourney V6.1?
Since it’s not really a new model but rather an update to V6, this new version focused on improving the pre-existing model’s creativity and image coherence. This means less missing (or extra) fingers, more realistic images, better creativity, and faster turnaround time. The team also put an emphasis on improving text generation, which we’re going to review in a bit.
Here’s a complete rundown of all the improvements they’ve done:
Since its release three weeks ago, Midjourney also released a new feature on their web app for V6.1 — a better editor. So, what exactly makes it better?
Back when Midjourney didn’t have a web app and can only be accessed on Discord, they already had a feature called region variation. This allows you to select an area of a generated image and Midjourney creates a new version of it. Take this image for example:
Let’s say I’m not really a big fan of the lake. Here’s how I’d change it using Midjourney’s Discord bot.
From there, you can select the region that you want changed.
Then, you get four new versions of your original image with the selected region changed.
So, what’s the problem?
The issue with this is that you don’t really have any control over the new images. You can’t change the prompt, so you’re stuck with the original prompt and Midjourney’s just going to keep generating new versions of the selected area using whatever prompt it already has.
With V6.1 and Midjourney’s web app, this is changed. Now you can select the area AND change the original prompt. So, for example, here’s a cute frog:
Let’s use the last one. When you click the image, a toolbar will appear on the right side. Select the button labeled “Editor.”
Let’s say I want to see its tongue. Weird request, I know. But what I can do now is press the “Erase” button above and select the mouth area. From there, I’ll just change the prompt from “A cute frog” to “A cute frog with his tongue sticking out.”
And now you don’t have to wonder anymore if Midjourney can edit photos. Here’s the final product:
Midjourney V6 vs. V6.1 Compared: Output Quality
Let’s compare Midjourney V6.1 to its previous version using the same prompts across different categories.
Portrait (People, Medium to Long Shots)
Prompt: portrait, an old wise man, glamour street medium format photography, shot on cinealta, day, cityscape background, vintage-inspired attire, soft ambient streetlights, reflective surfaces, subtle bokeh effect | |
Midjourney V6 | Midjourney V6.1 |
Both still give me a slight uncanny valley feeling, but V6.1’s version is a lot more coherent (as they promised). The older model still feels real if you don’t zoom in, but once you do, you can see slight issues with the man’s eye.
Portrait (People, Close-up Shots)
Prompt: somewhere in the year 1983, a woman looking up to the camera, birds eye view, melancholia, closeup, analog | |
Midjourney V6 | Midjourney V6.1 |
The face of the woman in Midjourney V6’s output is a little too smooth and brushed up for my liking. It also ignored the part of my prompt where I specified “bird’s eye view” angle.
The newer version did neither — showing better realism and prompt understanding.
Portrait (Non-Human)
Prompt: A hyperrealistic magnolia bonsai tree, studio lighting, dark background, isolated | |
Midjourney V6 | Midjourney V6.1 |
I think these are both good images, but if I were to nitpick, I’d say that the lighting on V6 went a little too overboard and it didn’t really generate much leaves compared to flowers.
Landscape
Prompt: victoria falls, leica m6, landscape photography | |
Midjourney V6 | |
Midjourney V6.1 |
While the colors of both images are muted to an extent, the former’s version is a little too muted for my liking. V6.1 has more balanced color grading.
Live Action Film Still
Prompt: a still from an early 2000s movie, a man in his bedroom reading a book, colorful background objects, whimsical elements, in the style of wes anderson, symmetrical | |
Midjourney V6 | |
Midjourney V6.1 |
Prompt: a still from a superhero movie, a superheroine who can control fire, street setting, in the style of marvel studios | |
Midjourney V6 | |
Midjourney V6.1 |
Prompt: a still from a 1960s japanese film, a samurai standing over a vanquised foe, in the style of akira kurosawa | |
Midjourney V6 | |
Midjourney V6.1 |
Beyond creativity, what I can say is that V6.1 follows the prompt more closely since their outputs are closer to the directors I mentioned. And what the hell is going on with those three feet from the first image?
Animated Film Still
Prompt: a ghost and a dog playing catch, whimsical, magical realism, in the style of hayao miyazaki, studio ghibli | |
Midjourney V6 | |
Midjourney V6.1 |
This one’s a matter of preference really — both images are good enough in my book. I do have to give the slight edge to Midjourney V6.1 because it feels more whimsy. But again, that’s just my opinion.
Digital Artworks
Prompt: pixel art scene, a sorceror’s library, 8-bit game | |
Midjourney V6 | Midjourney V6.1 |
Prompt: manila at twilight, digital art, low poly, geometric, colorful gradients, in soft color fields | |
Midjourney V6 | Midjourney V6.1 |
Prompt: an eldritch horror unleashed in the cosmos, dark fantasy illustration, lithography, 1980s illustration, gothic dark and macabre, larry elmore, lovecraftian | |
Midjourney V6 | Midjourney V6.1 |
The first one shows lower structural coherence in Midjourney V6 because, let’s be honest, who’s going to put a library outside their home? V6.1’s output also has more consistent pixel tiling compared to its older version’s inconsistent pixel sizes.
The second comparison shows more creativity in the part of Midjourney V6.1, in my opinion. It didn’t just take polygons and shaped them to look like a city — it’s much more intentional and uses color contrasts well.
And as for the third one, take your pick. Both are amazing.
Abstract Concepts
Prompt: a visualization of hope | |
Midjourney V6 | |
Midjourney V6.1 |
If I’m an art student, I’d feel like V6.1’s output is a lot more interpretable than V6 — which is just a bunch of glitter with bokeh in my opinion. I can’t see how it’s supposed to symbolize the concept of hope.
Logo (Without Text)
Prompt: a logo for a flower shop, clean pastel background, in the style of paul rand | |
Midjourney V6 | Midjourney V6.1 |
Prompt: a logo for a wine distillery, clean dark-ish background, minimalist, bauhaus style | |
Midjourney V6 | Midjourney V6.1 |
Prompt: a logo for a furniture store, art deco style | |
Midjourney V6 | Midjourney V6.1 |
While both are good, I think Midjourney V6’s outputs are a little too “loud” to be a logo. There’s just too many things going on.
Logo (With Text)
Prompt: a logo for a flower shop, clean pastel background, in the style of paul rand | |
Midjourney V6 | Midjourney V6.1 |
It’s the same issue as the one above: Midjourney V6’s output does too much to be used as a logo. As for text generation, both did really well to write “Austen’s” but the former version wrote it off-center — so I have to give the edge to V6.1.
Can You Use Midjourney For Free?
Unfortunately, no. If you want to use Midjourney, you need to avail of any one of their plans, which starts at $10 per month. Once you do, and let’s say you ran out of generations, you can rate images on their website and they’ll give you a free hour of generation in exchange for it.
What Else?
Midjourney is on a roll, and V6.1 is proof of this. Their latest update is a testament to the team’s dedication to pushing the boundaries of what’s possible in AI-generated art. And boy, I can’t wait to see what else is in store for the rest of 2024 since they promised V6.2 later this year.
The improvements in realism, coherence, and creative expression are flat-out impressive. Midjourney V6.1 is taking the visual storytelling game to a whole new level.
As the world of AI art continues to evolve at breakneck speed, it’s not DALL-E 3 or Firefly that’s leading the change, but Midjourney. With each iteration, they’re captivating and inspiring creatives everywhere, leaving us hungry for more.