Midjourney vs Adobe Firefly 2: Which Is Best? 12 Side-by-Side Examples
It's been a year since Adobe Firefly 2 was released. With better creativity and context understanding, can this AI image generator overtake its competition? We've spent quite a bit of time testing its limits and put Firefly to a direct comparison with Midjourney.
Justin Gluska
Updated September 23, 2024
Midjourney vs Adobe Firefly 2 Comparison of the Same Prompt
Reading Time: 8 minutes
Creating artwork is more accessible today than it has ever been — and that’s thanks to AI image generators. The power to create stunning and meaningful art just from a single prompt? Sign me up.
But not every text-to-image generator is created equally. Midjourney has been my go-to for the last few months and I’m a huge fan. However, I'd still say that it has some glaring weaknesses even with the 2023 release of V6 (and subsequently, V6.1 in 2024), particularly in nuance.
I’ve been testing it against other AI image generators as they all seem to keep improving and the next one on the list is Adobe Firefly 2. Released more than a year ago, this version claims to have better nuance and creativity than what the original version offered.
Just a quick disclaimer: there’s already a newer version of Adobe Firefly, and we’ve already compared it against Midjourney as well — so make sure to check that out.
Because of its native integration with other Adobe apps, some graphic designers have been giving Adobe Firefly a go and it’s safe to say that the reception has been positive. I've tried it in Photoshop and LOVE it. Makes my life so much easier.
But how do these tools stack up against each other?
A Bunch of Different Prompts Compared
To get a full scope of their creativity, I’ll be dividing the outputs into several categories including intent and prompt length. This will help us get a better idea on what areas they excel at and where they fail.
Before I start, you should also know that all Midjourney images are on the left and Firefly outputs are on the right.
Realistic Photo
Prompt: The Cliffs of Moher
Midjourney created a stunning artwork of the Cliffs of Moher during a sunset. The art is calm and vibrant, but it falls victim to the model’s tendency to overstylize its outputs. If we’re going by pure realism, then Firefly wins. It’s not as creative as Midjourney, but the image looks like a photograph. You could insert this into Google Images when I search the “Cliffs of Moher” and I wouldn’t bat an eye.
Close-Up
Prompt: An extreme close-up of a woman's face during a party
Firefly’s image has a “stock image” quality, which makes sense because most of the generator’s users will likely use it because of the Adobe ecosystem and integration to other apps such as Photoshop, After Effects, and more.
On the other hand, Midjourney’s output has a lot more texture and detail. You can see the imperfections on the woman’s face, and her eyes are a lot more detailed when you zoom in. I only wish that Midjourney didn’t ignore the “party” element in the prompt (like I said, nuance still could use some work), but overall, this is the clear winner.
Pastiche (Mimicking Other Artists)
Prompt: A quiet night in the suburbs, in the style of Van Gogh
Adobe Firefly lacks the imagination of a Van Gogh painting, which makes it look more like a children’s book illustration. Midjourney V6 completely understood the assignment. The sky borrows heavily from Van Gogh’s Starry Night and I like that the road reflects the light from the sky. No contest, Midjourney is way better at imitation.
3D Render
Prompt: A 3D render of a lit candle in a desk
This one’s tough. I genuinely like both of them. Midjourney leaned towards creating a realistic 3D render, while Firefly looks more like a painting. However, if I’m to use one of these two, I’d probably pick Midjourney because of the candle’s details. It also has more accurate lighting than Firefly.
Surrealist
Prompt: A surrealist painting of men who look like sheep walking in a busy Shibuya Crossing
I love Midjourney’s output here. It’s slightly unnerving, but in a good way. There’s a lot of detail in the background, which is important in establishing the Shibuya Crossing setting. It also helps that the artwork looks serious because the self-awareness adds a layer of surrealism.
Adobe Firefly has a more Dadaist take on surrealism. The sheep’s faces look pasted on people’s bodies. There’s no cohesion, but it still has that element of whimsy. However, I don’t like the softness of the background elements because it makes it look more AI than it needs to be.
Abstract Concepts
Prompt: Visualization of language
Asking AI image generators to create visualizations of abstract concepts helps us understand its nuance. Both Midjourney and Firefly did well here, in my opinion. The former interpreted language as knowledge within a person, whereas the latter understood it as a medium of communication in a community. This is a clear tie.
Film Still
Prompt: A film still of a kaiju taking down a building during the night
Firefly had a weird response to this prompt. Not only does the subject look nothing like a kaiju, it also created a still of an animation movie instead of a live-action one. There’s also no visible knocked down buildings. It didn’t understand the context at all.
Midjourney, however, created the perfect film still. It looks straight out of a Godzilla movie. It’s kinetic, action-packed, and realistic. This is one of my favorite outputs from Midjourney so far.
Architecture
Prompt: A two-storey bungalow in the middle of a quiet field
Midjourney has a more appealing output. The house is feasible and somewhere I’d actually love to live in. There’s also the little sneak peeks we get of the house’s interior, which contributes to its realism. Sell this house on Zillow and nobody would notice it’s AI.
I’d like you to zoom into Firefly’s house. Notice something off about the walls? The house doesn’t make sense. The walls are inconsistent and impossible. There’s also no window near the balcony and the house looks so… bland. It’s not a good architectural design at all, so that’s a point for Midjourney.
Flat Illustration
Prompt: A flat vector illustration of a cute Q-tip
Right off the bat, both these illustrations are pretty cute, but neither really looks like a q-tip. However, since the top looks like cotton, I’d pick Midjourney’s artwork for this category, especially since Firefly’s more reminiscent of a half-eaten popsicle stick.
Sketches
Prompt: A sketch of a rabbit jumping down a hole
This is another instance where Midjourney overstylized its output. The rabbit itself is good, but the rest looks chaotic. There’s no reason to add that pop of color to a sketch and it just contributes to the overall mess.
Sometimes, simplicity is key — and Firefly demonstrates that perfectly in this image. There are no unnecessary additions, jump a simple rabbit jumping out (or rather, into) a hole. It also feels more “human” than the Midjourney one because it captures freehand artwork better.
Text Generation
Prompt: The menu of a rustic coffee shop
With DALL-E 3 promising better text generation, I wondered how well these two can create text in images.
Any version of Midjourney lower than V6 will falter when asked to generate text. Fortunately, they already fixed that issue. It’s still far from perfect (and I’d argue that DALL-E 3 is still the better option, especially with GPT-4o) but it’s way better than Firefly 2, which just ended up looking like a bunch of gibberish.
High Context
Prompt: A weary man wearing a blue shirt and black slacks protecting his two daughters and one son against falling asteroids in New York City during a white Christmas night.
I’ve been using Midjourney for months and one thing’s for certain: it sucks at high-context prompts. Or at least, it used to be.
Like I said earlier, V6 was a huge leap in nuance and features for Midjourney. It didn’t miss any single element of my prompt. The outfit’s there, all the children’s there, even the asteroids are there.
As for Firefly, it not only missed several elements of my prompt, it also missed the mark completely of the prompt’s intent. The family in the photo looks happy and unbothered, and there are no impending meteorites in the sky. This is a clear win for Midjourney.
The Final Verdict
I’d still use Midjourney over Firefly. It’s more robust at image generation and, with V6.2 around the corner, we can expect Midjourney to take that leap that would push them way above its counterparts. There’s no telling what this model can do.
Overall, Midjourney is simply more creative than Firefly. It’s better at creating artwork and understanding context. That said, Firefly is still a good alternative if you’re looking to create more realistic photos or you’re a graphic designer who’s using the Adobe ecosystem.
The best AI image generator is really just the one that fits your style better. Use this as a guide when you're on the edge of finding your next subscription.
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.