Comparing DALL-E and Midjourney AI Images in 2024

I have previously compared DALL-E and Midjourney back in 2022, when the technology was still a bit rough and ready. Since both tools have come a long way, I figured it was worth comparing them again in the first quarter of 2024 to see how they stacked up.

A bit about DALL-E and Midjourney

  • Both are online tools for generating online images.
  • DALL-E was created by Open AI and is used from within the ChatGPT interface.
  • Midjourney is an AI tool which began as more abstract, stylized and beautiful than other tools.
  • The current state of things seems to be that DALL-E can more faithfully recreate complex prompts, while Midjourney images look far better and more realistic.

Prompt 1: A chipmunk businessman relaxing in a first class seat on a plane.

Dall-E

DALL-E produces two images, both very similar. And they clearly match the prompt. It’s a chipmunk, undoubtedly a businessman. And he’s obviously in first class.

Midjourney

Midjourney offers an interesting mix. There are some images that show a business chipmunk in first class. But it’s not as seamless as DALL-E. I think the image I’ve highlighted below encapsulates the main strength of Midjourney, in that while it doesn’t match the prompt (it’s a squirrel, not a chipmunk, and there’s no sign he’s a businessman) its very realistic and you could mistake it for an actual photograph,

Prompt 2: a cowboy stares out over a beautiful sunset. Western movie. Cinematic. 35mm film. film grain.

Dall-E

a cowboy stares out over a beautiful sunset. Western movie. Cinematic. 35mm film. film grain. DALLE

DALL-E struggles with matching film styles well. The images are very pretty and match the description in the prompt of WHAT is happening but is unable to successfully replicated an image from a movie.

a cowboy stares out over a beautiful sunset. Western movie. Cinematic. 35mm film. film grain. DALLE

Midjourney

a cowboy stares out over a beautiful sunset. Western movie. Cinematic. 35mm film. film grain. midjourney

Midjourney on the other hand is able to capture the TYPE of image I wanted. It’s clearly a western and the image below could easily have been taken from a movie.

a cowboy stares out over a beautiful sunset. Western movie. Cinematic. 35mm film. film grain. midjourney

Prompt 3: a Japanese woman wearing a blue baseball cap, standing in front of the pyramids, holding up a sign that says AI

Dall-E

a Japanese woman wearing a blue baseball cap, standing in front of the pyramids, holding up a sign that says AI DALLE 2

This example I think captures the strengths and weaknesses of the DALL-E currently. It is spot on in matching the prompt. Every detail is accounted for. But it’s clearly fake and obviously AI generated.

a Japanese woman wearing a blue baseball cap, standing in front of the pyramids, holding up a sign that says AI DALLE image

Midjourney

a Japanese woman wearing a blue baseball cap, standing in front of the pyramids, holding up a sign that says AI midjourney AI

This prompt also captures the strengths and weaknesses of Midjourney. Comparing DALL-E with Midjourney, Midjourney gets the details almost correct, but AI is written on her blue cap and not on the sign. DALL-E gets the prompt right, but it looks bad. Midjourney gets it wrong but it looks amazing. Almost photo realistic.

a Japanese woman wearing a blue baseball cap, standing in front of the pyramids, holding up a sign that says AI midjourney

Prompt 4: Friends enjoying a braai in South Africa

Dall-E

Friends enjoying a braai in South Africa DALLE Comparing DALL-E and Midjourney

Once again DALL-E gets the details correct, but the style is a certain “DALL-E” style. Accurate but clearly AI generated. One image has the amusing but incorrect assumption that braais would be placed on a table. Perhaps mixing up a braai with a Korean BBQ?

Friends enjoying a braai in South Africa DALLE

Midjourney

Comparing DALL-E and Midjourney Friends enjoying a braai in South Africa Midjourney AI image

Midjourney produces images that are definitely closer to photos than DALL-E. In some images it also shows a braai on a table which I found quite funny.

Friends enjoying a braai in South Africa midjourney AI

Prompt 5: Robocop anime in the style of Hayao Miyazaki

Dall-E

Here we get the first major difference between DALL-E and Midjourney. DALL-E won’t created the image as it replicates the style of an artist who’s work was created after 1912.

When asked about the 1012 cutoff, DALL-E / ChatGPT / OpenAI responded with the following;

“The date 1912 serves as a cutoff point in our content policy to distinguish between public domain and copyrighted works. Typically, works created by artists who were active or whose works were published prior to 1912 are considered to be in the public domain in many jurisdictions. This means they are no longer under copyright protection and can be freely used without infringing on copyright laws. For artists, creative professionals, or studios whose latest work was created after 1912, their styles, characters, and creations are likely still under copyright protection. This guideline helps ensure respect for intellectual property rights and avoids the generation of content that could infringe on the copyrights of modern artists and creators.”

Midjourney

Robocop anime in the style of Hayao Miyazaki

Midjourney on the other hand had no problem with creating images of a Robocop anime in the style of Hayao Miyazaki, the Oscar winning Japanese animation legend.

Midjourney on the other hand had no problem with creating images of a Robocop anime in the style of Hayao Miyazaki, the Oscar winning Japanese animation legend.

My Other Blog Posts

Hope you found this blog about comparing DALL-E and Midjourney interesting. If you enjoyed this you might like my other blog posts.

Comparing DALL-E and Midjourney AI Images in 2024
Four in One Analytics Looker Studio Dashboard Template
Visualizing a young Africa and an aging World
Visualizing Fertility Rates across the World
GDP per Capita within African Regions