Dall-E 3 vs. Midjourney: A Big Comparison of the Most Advanced AI Art Generators

Share this article

• 
• 
• 
• 
• 

by Damir Yalalov

Published: October 02, 2023 at 2:52 pm Updated: October 03, 2023 at 3:11 am

by Danil Myakin

Edited and fact-checked: October 02, 2023 at 2:52 pm

To improve your local-language experience, sometimes we employ an auto-translation plugin. Please note auto-translation may not be accurate, so read original article for precise information.

Join us on this thrilling journey as we explore Dall-E 3 and Midjourney’s subtleties, complexities, and untapped potential. This article highlights the most intriguing comparisons based on research done by AI enthusiast Atachkina; if you’re interested in learning more, click the link.

This article provides a text-to-image prompt, an image showing the results from Dall-E 3 and Midjourney, and an explanation of the differences between the two art generators. Let’s begin.

Both neural networks performed admirably in this case, with the Midjourney slightly outperforming the others.

Dall-E 3 did a much worse job here; it got the bright colours of the styles, but not the clarity of the details; deformed bodies appeared in the background, and the faces were not at all successful.

It turned out to be interesting both places, but Dall-E 3 once more struggled with the faces. Instead, it made a plush beige bag as instructed in the prompt, and Midjourney disregarded it. In this instance, Dall-E 3 was very obedient in carrying out the prompt.

And once more, while both grids make excellent collages, Dall-E 3 is more faithful to the prompt; it added only the heroes we specified, it couldn’t turn into a joker, and it crossed the captain with Batman.

Midjourney was able to combine the two artists’ respective styles from the prompt, whereas Dall-E 3 just added a lot of busy details and bright colours to the background.

Once more, the cats are in top form, and both neural networks comprehend film cameras perfectly. However, Dall-E 3 even adds grain to the pictures.

Dall-E 3 created a young Leonardo DiCaprio with cool jumper textures, added film grain and colour scheme and very coolly reflected the feel of a Russian dacha. Midjourney was a good colour reflector for the movie, and DiCaprio gave her a more mature appearance.

Although both neural networks are adept at creating collages, if you look closely, Midjourney distorts faces and some object shapes, while Dall-E 3 is more accurate in the execution of the characters themselves—it even turned out to be Chewbacca.

When you zoom in on the photographs, you’ll notice that Dall-E 3 has blurry eyes; Midjourney, on the other hand, is flawless. Dall-E 3 also prescribed a brand; the snakes on the heads appear to be more alive and in motion; Midjourney always made them lying down, rather than on the head.

Both are cool, but Midjourney considered the artist’s style as well as the effect of a film camera, whereas Dall-E 3 ignored the full-length shot and did not consider it.

We also made the decision to test a photo with fairies, but Dall-E 3 obstinately refused to cooperate. Midjourney did not ignore the wings because the reference with wings had been added. When Dall-E 3 did take a picture, it offered some intriguing possibilities, but with an American woman.

Midjourney did a fantastic job, but we want to draw special attention to how Dall-E 3 created the film effects in the top right picture and added own white handwriting; it turned out great.

Dall-E 3 was able to very obediently realise all the heroes of the prompt in one image once more. Midjourney tried very hard and even came close to succeeding.

At first glance, it appears that both are good, but closer inspection reveals that the Dall-E 3 lacks photorealistic volume and that Midjourney handled the joints with forks with a bang.

Both generators are proficient in their respective fields, with Dall-E 3 excelling in text and Midjourney excelling in photorealism.

The physics and geometry of hair dryers are difficult for Midjourney. You can spend a lot of time struggling with tries and references, and occasionally the results resemble a hair dryer, but Dall-E 3 produced an acceptable result on the first try and even wrote the text.

The only eye is good, but that’s another story. In Midjourney, we wrote a negative prompt – no cartoon, illustration, flat, two eyes. Dall-E 3 immediately obeyed and made one eye, a smile, and a hat off, but it flatly refused to let anyone take her picture.

Midjourney made the generation not like Brad, so we used the extra service Insight Face Swap to put Brad’s face on the generation; there was a post about it here. Dall-E 3 knows who Brad Pitt is and can draw stars without any additional software.

Both meshes are good, but Dall-E 3 can create unicorn horns while Midjourney typically cannot.

Dall-E 3 did a good job of putting the characters into action; we can see an orc and an elf with elf ears. There is also a person wearing a Nike tracksuit, but their eyes are smudged. The elven pointed ears are mostly ignored by Midjourney, and Nike is also disregarded.

When the postscript “illustration” was initially left out of the prompt, Dall-E 3 created one. We then decided to compare it to Midjourney’s illustration. While Midjourney more closely resembled Soviet-era illustrations and did not include the fairy wings, Dall-E 3 did a fantastic job drawing the hammer and sickle. The example to the right shows how Dall-E 3 might appear in the text.

However, Midjourney went into photorealism; there is no main character in the images, only the surroundings, but still cool. Dall-E 3 didn’t want to be in the photo again.

Dall-E 3 vs. Midjourney: Pros and Cons

As users explore this technology, several notable strengths and limitations have come to light, shedding further insight into its functionality.

Pros:

1. Prompt Obedience: One of the standout features of Dall-E 3 is its remarkable ability to follow prompts accurately. Users have reported that the AI model responds effectively to a wide range of input, making it a versatile tool for various tasks.
2. Multifaceted Creativity: Dall-E 3 exhibits the capability to depict multiple characters within a single image, expanding its potential for storytelling and creative projects. This multifaceted approach enhances its utility across different domains.
3. Text Integration: Users have noted Dall-E 3’s proficiency in integrating text seamlessly into images. This feature facilitates the creation of visually engaging content with embedded textual elements.

Cons:

1. Image Clarity: A notable limitation is the AI’s tendency to produce images with blurred faces and eyes. While it excels in creativity, it sometimes lacks the clarity and precision seen in human-generated content.
2. Style Consistency: Dall-E 3 doesn’t consistently replicate specific artists’ styles, which may be a drawback for those seeking precise artistic emulation.
3. VPN Requirement: Access to Dall-E 3 currently necessitates the use of a VPN, which may pose accessibility challenges for some users.
4. Image Management: Users have encountered limitations when managing generated images on the Microsoft Bing website. Notably, there’s no format orientation function, and image history is restricted to recent uploads, necessitating immediate copying for later use.
5. Generation Speed: In some cases, the generation process in Dall-E 3 has been reported to be slower compared to other AI models.

Despite these limitations, Dall-E 3 holds substantial promise. Users and experts alike recognize its potential to revolutionize content creation and storytelling. As OpenAI continues to refine and expand its offerings, it’s expected that Dall-E 3’s strengths will shine even brighter, making it a valuable tool in various fields.

FAQs

Both Dall-E 3 and Midjourney have their strengths and weaknesses. Dall-E 3 is notably obedient to prompts and can integrate text seamlessly into images. However, it sometimes produces images with blurred faces and eyes and may not consistently replicate specific artists’ styles. On the other hand, Midjourney excels in photorealism but may not always capture the essence of certain prompts as accurately as Dall-E 3.

The article provides text-to-image prompts, showcasing the results from both Dall-E 3 and Midjourney, and explains the differences between the two art generators.

Both AI models have their strengths and weaknesses. For instance, in a prompt about a spaceman on Jupiter, Midjourney slightly outperformed Dall-E 3. However, in another prompt about Wonder Woman, Dall-E 3 was more accurate in capturing the essence of the prompt.

• Prompt Obedience: Dall-E 3 accurately follows prompts.
• Multifaceted Creativity: It can depict multiple characters in a single image.
• Text Integration: Dall-E 3 can seamlessly integrate text into images.

• Image Clarity: It sometimes produces images with blurred faces and eyes.
• Style Consistency: Dall-E 3 doesn’t consistently replicate specific artists’ styles.
• Image Management: There are limitations when managing generated images on the Microsoft Bing website.
• Generation Speed: Dall-E 3’s generation process can be slower compared to other AI models.

About The Author

Damir is the team leader, product manager, and editor at Metaverse Post, covering topics such as AI/ML, AGI, LLMs, Metaverse, and Web3-related fields. His articles attract a massive audience of over a million users every month. He appears to be an expert with 10 years of experience in SEO and digital marketing. Damir has been mentioned in Mashable, Wired, Cointelegraph, The New Yorker, Inside.com, Entrepreneur, BeInCrypto, and other publications. He travels between the UAE, Turkey, Russia, and the CIS as a digital nomad. Damir earned a bachelor's degree in physics, which he believes has given him the critical thinking skills needed to be successful in the ever-changing landscape of the internet. 

More articles

Damir Yalalov

• 
• 

Damir Yalalov

Damir is the team leader, product manager, and editor at Metaverse Post, covering topics such as AI/ML, AGI, LLMs, Metaverse, and Web3-related fields. His articles attract a massive audience of over a million users every month. He appears to be an expert with 10 years of experience in SEO and digital marketing. Damir has been mentioned in Mashable, Wired, Cointelegraph, The New Yorker, Inside.com, Entrepreneur, BeInCrypto, and other publications. He travels between the UAE, Turkey, Russia, and the CIS as a digital nomad. Damir earned a bachelor's degree in physics, which he believes has given him the critical thinking skills needed to be successful in the ever-changing landscape of the internet. 

• 

More articles