Flux.1 vs. Midjourney: Now There’s Competition

Flux.1 is the most promising AI image generation model I’ve seen since Stable Diffusion. So, let’s see how well it does against the final boss: Midjourney.

John Angelo Yap

Updated August 22, 2024

Visualized battle between AI image generators, generated with Midjourney

Visualized battle between AI image generators, generated with Midjourney

Reading Time: 7 minutes

AI image generation is evolving at breakneck speed. Just when you think you've seen it all, something new blows your mind.

Remember when DALL-E and Midjourney were released? It felt like pure magic. Suddenly, anyone could create stunning AI-generated art with just a few words. But the world of artificial intelligence doesn't stand still for long. More and more models were created by the minute, like Firefly and Meta.

And now, we have Flux.1, the latest AI image generator. Created by some of the brilliant minds behind Stable Diffusion, this newcomer is now the talk of the town on Twitter. Its promising photorealism could give Midjourney a run for its money. The buzz around Flux.1 vs Midjourney is real.

So how do these two AI powerhouses really stack up? What are their strengths and weaknesses? Let's find out.

What are Luma AI and Midjourney?

Let’s start with Midjourney, since that’s the one people are most familiar with. This AI image generator has been around since 2021, and it’s probably the most popular, with more than 20 million users as of August 2024. Midjourney takes a prompt and creates an image based on it — and this is just one of its many features, which we’ll get to in a bit.

Here’s our full review of Midjourney. If you’re also interested in prompting, here’s a great article to start with. 

So, Flux.1. It’s more similar to Stable Diffusion than Midjourney since, to be fair, it’s created by some of the people who helped develop Stable Diffusion. It’s an open-source model that can create images from prompts. From what I’ve been seeing online, Flux.1 looks to be a promising tool for generating realistic images — something that Midjourney and DALL-E struggle with.

Flux has three different models: Flux.1 [pro] for commercial use, Flux.1 [dev] for a more distilled version of the previous model, and Flux.1 [schnell] for personal use.

Flux.1 vs. Midjourney: Features

Midjourney is just a beast when it comes to features. It has common features like panning, zooming, image-to-prompt, and variations. But on top of that, they also have more unique features like style reference, character reference, and upscaling. And those are just the tip of the iceberg.

Since releasing V6.1 and their web interface, Midjourney also upgraded their region variation feature to work more like an editor. It now allows you to highlight an area from an image using a brush and change the prompt, affecting only the selected region.

Unlike Midjourney, Flux.1 doesn’t really have an interface since it’s just a model and not an application. While this makes it a bit more inaccessible to beginners, this also makes it a much more versatile option, especially for those who are already familiar with Stable Diffusion. 

That said, some of the websites that offer Flux.1 do offer you some control over aspect ratio, number of variations, output format, image quality, and NSFW enabling. For example, take this Replicate website that’s hosting Flux.1 [schnell]:

Overall, I’d still say Midjourney has more features built on top of it, so it takes the cake. But if we’re only talking about potential, then Flux.1 might just be the next big thing.

Oh and before I move onto the next part, I’d just like to point out one specific use case for Flux.1 that I found online: you can train it with your own image. What does this mean? Well, just look at this:

Flux.1 vs. Midjourney: Output Quality

Again, Flux.1 doesn’t have an interface. So, for this test, I’m going to use Replicate’s Flux.1 [schnell] image generator.

Realism Photos

Prompt: portrait, a blonde woman walking along a shore, candid photography, analog, shot on cinealta
Flux.1 [schnell]Midjourney V6.1
Prompt: portrait, a blonde woman walking along a shore, candid photography, analog, shot on cinealta
Flux.1 [schnell]Midjourney V6.1
Prompt: a young professor teaching physics
Flux.1 [schnell]Midjourney V6.1

Before writing this article, I did my due diligence by looking for Flux.1 generated images on Twitter. What I’ve seen set the bar high for my expectations.

But unfortunately, those expectations haven’t been met.

Don’t get me wrong — Flux.1 is still amazing for realism. It’s just that Midjourney’s better than base Flux.1 (at least, against its base [schnell] model). Images from Midjourney actually look like they’re taken by a person, and not generated by an AI. Flux.1 also didn’t follow the prompt I gave for the second comparison.

That said, I have no doubt that, once the model’s fine-tuned, Flux.1 can be just as good (if not better) than Midjourney.

As for Midjourney’s images, I don’t have any complaints except for the last image. Seems like there are still some issues here and there with generating faces even using their latest model.

Landscape Photos

Prompt: the grand canyon, landscape photography 
Flux.1 [schnell]
Midjourney V6.1
Prompt: aurora borealis 
Flux.1 [schnell]
Midjourney V6.1

Flux.1’s outputs seem a little more smooth and brushed up. At first glance, they appear to be real images, but once you zoom in, you can quickly identify it as AI-generated. Midjourney has the same problem to an extent, but it’s a bit less noticeable in my opinion.

Digital Art

Prompt: pixel art, a view of the ocean from the shore during sunset
Flux.1 [schnell]Midjourney V6.1
Prompt: vector art, a kid who can control fire and ice, 80s comics style
Flux.1 [schnell]Midjourney V6.1
Prompt: beautiful anime artwork, a schoolgirl turning into an angel, neon streets, anime screencap style
Flux.1 [schnell]Midjourney V6.1

For the pixel art, I actually prefer Flux.1’s simplicity and particularly like how accurate the pixels are. The pixels on Midjourney’s version bleed into each other.

As for the other two, there’s no question — Midjourney’s creativity outshined Flux.1.

Prompt: a minimalist logo for a barbershop
Flux.1 [schnell]Midjourney V6.1
Prompt: a chair icon
Flux.1 [schnell]Midjourney V6.1

For the first set of logos, I prefer Flux.1’s output because it has more of a mass appeal than Midjourney’s version. But for the second one, Midjourney takes the cake for creating a logo that’s more unique than Flux.1’s version.

Text Generation

Prompt: a convenience store named “12AM SNACKS” 
Flux.1 [schnell]
Midjourney V6.1

Well, both did pretty great here at writing the name of the store, so no complaints from me. It’s really great to see how far AI image generators have come in just a short time when it comes to text generation.

Flux.1 vs. Midjourney: Price

Flux.1 is technically free, but that comes with a few asterisks. Because of its many (and by many, I mean 12 million) parameters, you need a pretty expensive GPU along with a powerhouse system to run it. So, the easiest way to use it is to access it via their partner sites like Hugging Face, Mystic, and Replicate. 

Some of those I mentioned allows you to use Flux.1 [dev] and Flux.1 [schnell] for free until you hit a certain number of runs.

On the other hand, Midjourney is not free. You need to pay for one of their plans, which starts at $8 per month. 

The Bottom Line

So, what's the final verdict in this clash of the AI image generators? For now, Midjourney still wears the crown for general images, but Flux is a lot better for people and faces. Its new and already-polished interface, wealth of features, and consistently impressive AI art outputs keep it at the top of the pack. And this isn't the end of the road for Midjourney too, since it's only going to keep getting better from here.

But don't count Flux.1 out just yet. This scrappy newcomer shows incredible promise, especially in the realm of photorealistic AI-generated photos. The race between Flux.1 and Midjourney is far from over.

So, who’s the real winner in all of this? That's us — the users. This fierce competition between AI image models is driving rapid innovation. What seemed impossible just years ago is now at our fingertips, and the ceiling for AI art creation keeps getting higher.

Want to Learn Even More?

If you enjoyed this article, subscribe to our free newsletter where we share tips & tricks on how to use tech & AI to grow and optimize your business, career, and life.


Written by John Angelo Yap

Hi, I'm Angelo. I'm currently an undergraduate student studying Software Engineering. Now, you might be wondering, what is a computer science student doing writing for Gold Penguin? I took up studying computer science because it was practical and because I was good at it. But, if I had the chance, I'd be writing for a career. Building worlds and adjectivizing nouns for no other reason other than they sound good. And that's why I'm here.

Subscribe
Notify of
guest

0 Comments
Most Voted
Newest Oldest
Inline Feedbacks
View all comments