Midjourney vs. DALL·E 3: These Image Generators AI Tools, DALLE 3 and Midjourney are two of the best AI art generators available today. Regardless of how strange or unusual your request is, both can take a text prompt and generate a series of matching images. DALLE 3 is probably the easiest to learn, while Midjourney, despite being a little difficult to use, has some of the most powerful features.
I’ve been testing both of these image generators professionally and personally since their release, and there’s a lot to unpack. So let us begin.
How do DALL·E 3 and Midjourney work?
DALLE 3 and Midjourney were both trained on millions or billions of text-image pairs, allowing them to understand concepts such as dogs, deerstalker hats, and dark moody lighting. This is how they can decipher what a prompt like “an impressionist oil painting of a Canadian man riding a moose through a forest of maple trees” asks of them.
When it comes to creating images, they employ a technique known as diffusion. They begin with a random field of noise and then edit it in a series of steps to better match their interpretation of your prompt. This is why you can get different results every time you try the same prompt: the randomness of the starting seed can completely change the outcome. In a previous article comparing DALLE 3 and Stable Diffusion, I described the process as similar to looking up at a cloudy sky, spotting a cloud that resembles a dog, and then being able to snap your fingers to make it more and more dog-like. While there is more to it, it’s not a bad way to think about things.
Of course, just because both models use the same image-generation technique doesn’t mean you’ll get similar results from DALLE and Midjourney. How each model interprets your prompt, the weight it assigns to various parameters, the data they were trained on, and the philosophies of the companies that created them all have a significant impact on the output.
DALLE 3’s interpretation of “an impressionist oil painting of a Canadian man riding a moose through a forest of maple trees.”
And here’s Midjourney’s.
As you can see, DALLE 3 has done a much better job of depicting my Canadian man riding a moose, though Midjourney’s output looks far more like a genuine impressionist painting.
I’ll go into more detail about these differences as we go, but don’t declare either AI model a winner just yet.
DALL·E 3 vs. Midjourney at a glance
Both DALLE 3 and Midjourney do similar things, but there are some significant differences. Here’s a quick rundown of the main distinctions, but keep reading for a more detailed breakdown.
| DALL·E 3 | Midjourney | |
|---|---|---|
| Quality | ⭐⭐⭐⭐⭐ Exceptional AI-generated images | ⭐⭐⭐⭐ Exceptional AI-generated images, but may occasionally miss key bits of a prompt |
| Ease of use | ⭐⭐⭐⭐⭐ Collaborate with a chatbot | ⭐⭐ Awkward at every step |
| Power and control | ⭐ Very limited options beyond asking the chatbot to rerun a slightly different prompt | ⭐⭐⭐⭐⭐ Best-in-class prompting and editing options |
DALL·E 3 is easier to use
DALLE 3 is accessible via ChatGPT, Bing Image Creator, Microsoft Paint, and other services that use its API. All provide a broadly similar experience, though ChatGPT is the most official and easiest to use, so I’ll concentrate on it for this comparison. You must be a ChatGPT Plus subscriber to access it, so if you aren’t, consider one of the other free options.
Even though there are several ways to access DALLE 3, they all function similarly and are incredibly simple: you simply ask DALLE 3 to create an image for you.
Midjourney, on the other hand, is awkward from the beginning. It does not have a web application. You must instead use Discord, a team chat app, to get access to it. It’s not difficult to create a Discord account and join Midjourney’s server, but it’s an unusual way to access an AI art generator.
This also applies to using Midjourney. Instead of a logical interface, you must prompt the Midjourney Discord Bot by typing /imagine /your prompt>—either in a direct message or in a public channel. That’s before we even get into the arcane parameters and various remixing features you can use to better control the results.
So, if you just want to try out an image generator, DALLE 3 is a lot easier to use.
Midjourney is a lot more powerful
Despite its awkwardness, Midjourney has a lot to offer. It’s probably the most powerful AI image generator available right now, assuming you’re not willing to train your own custom model.
Let us begin with the parameters. There is a full list here, but by typing -- followed by the relevant command, you can control things like the aspect ratio of your images, how varied they are, what seed to start with, and even whether you want to create repeating tiles for seamless patterns.
Then there are the tools for enlarging, varying, and editing your images. After you run a prompt, you’ll see two sets of four buttons: U1, U2, U3, and U4, which upscale the corresponding image to 1024px wide, and V1, V2, V3, and V4, which rerun your prompt to create more variations based on the corresponding image.
Once you’ve upscaled an image, you can use the Upscale (2x) or Upscale (4x) buttons to enlarge it to 2048px wide—or even 4096px wide. You can make more variations that either closely match the starting image or are more distinct with Vary (Strong). You can even use Vary (Region) to change specific parts of the image.
Furthermore, the Zoom Out and Pan (the arrows) buttons allow you to extend your creation’s boundaries. When you activate Remix mode, you gain even more control because you can change your prompt each time you make a variation.
That’s before we even get into features like image prompts or blend, which let you combine the “concepts and aesthetics” of multiple images you upload.
DALLE 3 comes nowhere near matching this feature set. While you can change the aspect ratio, everything else is dependent on getting ChatGPT to rerun a slightly different prompt. Even the Bing tools that use DALLE 3 don’t provide many additional options. The only ones worth mentioning are Image Creator, which allows you to import your image directly into Microsoft Designer, and Paint, which allows you to generate images in the app and edit them (or at least paint over them).
When you use ChatGPT to request changes, it simply generates new images with a new prompt.
Surprisingly, at least for the time being, DALLE 2 gets closer to Midjourney. It at least allows you to inpaint (change the internal parts of an image) and outpaint (add content outside of the image boundaries). The model is far from perfect, but you have more control.
Both make great AI-generated images
Despite this, both DALLE 3 and Midjourney are capable of producing stunning images. Although DALLE 3 is faster and easier to use, you have less control over the results. Midjourney can be a little obtuse at times, but you have a lot more control over how things appear.
Nonetheless, there are a few distinctions worth noting. DALLE 3, because it integrates with ChatGPT, does an excellent job of interpreting prompts. It appears to be more capable of handling both shorter and longer, more complicated prompts.
Midjourney doesn’t make mistakes very often, but it takes a lot more steering. Instead of writing long descriptions, I discovered that it was more effective to give it a series of key words.
Also, because it has good editing tools, I was more willing to accept a less-than-ideal first image and then use remixing, variations, and other tools to get a final image I liked.
Pricing
Pricing depends on your needs
DALLE 3’s pricing is straightforward: it costs $20 per month as part of ChatGPT Plus, or it’s free as part of various Microsoft tools, though some of them will watermark your images. There are no published limits on how many images you can generate each day or month with DALLE 3, but those limits are presumably there to prevent people from creating images indefinitely.
Midjourney does not offer a free option, but the Basic Plan starts at $10/month and includes 200 minutes of GPU time. Of course, this is where things get complicated. According to Midjourney, that’s good for about 200 generations per month, but it all depends on what you’re asking it to do. If you make a lot of variations and upscale them all to the maximum, you’ll burn through those GPU hours much faster than if you make a lot of low-res images.
To complicate matters further, starting with the $30/month Standard plan, you get more fast GPU hours, but you can generate an unlimited number of images in Relax mode—which only runs when there is free GPU power.
Given all of the extra features that both apps provide, I’m hesitant to make any price comparisons between DALLE 3 and Midjourney. If the $10/month Midjourney plan meets your needs, it is probably the best balance of features and price, but you can use DALLE 3 for free through the Bing Image Creator, or for $20/month as part of ChatGPT Plus—which also includes all of ChatGPT’s language generation features.
Commercial use is complicated
Things become more complicated if you intend to use DALLE 3 or Midjourney for commercial purposes. Both models permit commercial use (though not if DALLE 3 is used through Microsoft), but the full legal implications have yet to be explored.
The United States Copyright Office ruled in February 2023 that images created by Midjourney, and thus other generative AIs, cannot be copyrighted. This means you have few safeguards if someone steals your photos and uses them in ways you don’t want them to. Using someone else’s image technically violates Midjourney’s terms of service, but that’s not exactly a strong legal defense if you’re trying to build a brand or create character designs with the app. Midjourney’s worst-case scenario is that they will ban whoever takes your photos.
Technically, I’d probably recommend Midjourney if you want to monetize your AI creations, simply because its model gives you more flexibility. DALLE 3 will completely prevent you from creating a massive amount of content, including images of public figures.
Midjourney is weird
Because I’ve been attempting to compare DALLE 3 and Midjourney as closely and rationally as possible, there are some odd and awkward aspects of Midjourney that I haven’t been able to address. Three major points to consider are:
- Unless you subscribe to the $60/month Pro plan and enable Stealth Mode, all of your images are automatically published to Midjourney’s member gallery, where anyone can view, download, and copy your prompts.
- Every setting and feature is managed via Discord bot commands. It will always be strange.
- Midjourney has an anime-specific model called Niji, which you can enable by adding –niji 5 to the end of a prompt or by typing /settings and selecting it.
The help docs are excellent, and I’ve never had any major issues with Midjourney, but I can’t emphasize how strange it feels to use if you’re expecting a more typical app. They’re apparently working on a regular web app, so I’m looking forward to seeing it.
DALL·E vs. Midjourney: Which should you use?
Most people should be able to choose between DALLE 3 and Midjourney fairly easily:
- Midjourney is a great choice if you want the best AI image generator currently available, the ability to tweak and edit your images, and are willing to work around its many quirks.
- DALLE 3 is the simple solution that consistently produces excellent results.
Alternatively, you could investigate Stable Diffusion. It’s less eccentric than Midjourney but more potent than DALLE 3.
[…] Midjourney vs. DALL·E 3: Which image generator is better […]