AI Showdown: MidJourney vs DALL-E 3 - Unleashing the Power of Prompts!

AI image generation technology is advancing at such a swift pace that within a matter of weeks or months, the possibilities in terms of quality and features can be completely transformed. With the introduction of DALL-E 3, a significant technological breakthrough occurs, but how does it compare to MidJourney?

What's Special About DALL-E 3?

We have previously extensively covered the evolution and capabilities of MidJourney. It has been the preferred choice for generating high-quality artistic images suited for practical applications. However, achieving the desired image in MidJourney can be a hit-or-miss experience. For precise control, one would need to turn to Stable Diffusion along with its various mods like ControlNet. But using Stable Diffusion is considerably more challenging, whereas both MidJourney and DALL-E 3 excel in terms of user-friendliness.

DALL-E is designed to adhere more closely to the specific instructions provided. In other words, if you request particular character poses, scene details, or object arrangements, DALL-E 3 should, in theory, produce the desired outcome. We will now compare DALL-E 3 and MidJourney using multiple prompts. Each AI generator will receive the same prompt.

Prompt 1: Artistic Flair

First, I just want to get a general feel for what each generator will do artistically, so we'll start with a rather generic prompt:

Generate an image of an epic fantasy scene with elves and dragons in a 90s fantasy art style

Here's the MidJourney image I thought was best.

And here's the DALL-E 3 image I thought was best.

It's worth noting that the image generator is not directly influenced by my specific prompt when using ChatGPT (the interface for DALL-E 3). One of the key features of DALL-E 3 is its utilization of ChatGPT (i.e., GPT-4) to handle the "prompt engineering" aspect on your behalf. Consequently, it generates more intricate prompts to enhance the quality of outcomes. Below is the prompt ChatGPT formulated based on my inquiry:

An oil painting evoking the whimsical fantasy art of the 90s, featuring a party of male and female elves poised on the precipice of a cliff. Majestic dragons dominate the sky, their wings casting ethereal shadows over a flourishing forest below. The artwork is abound with vivid hues and captivating lighting, creating a truly immersive experience.

We now have two options that can be compared. But which one is the winner? In my view, the DALL-E 3 image is closer to my original request, whereas the MidJourney image possesses a more distinct style and a stronger artistic touch. Personally, I find MidJourney's current V5 model particularly impressive in terms of its overall artistic flair, but it should be noted that this is a highly subjective judgment.

For the remaining comparisons, I will exclusively use the GPT-generated prompts for both image generators, in order to eliminate any influence stemming from my own skills (or lack thereof) in crafting prompts. In simpler terms, I will first request an image from ChatGPT, and then copy and paste the most impressive prompt it generates into MidJourney.

Prompt 2: Text Elements

MidJourney often produces nonsensical text in generated images, which means that T-shirts with text or store signs will not display any coherent information. However, DALL-E 3 guarantees the ability to accurately place and create any desired text within the image frame. Let's put this claim to the test using the following prompt provided by ChatGPT:

The image depicts a computer geek fully immersed in coding, resembling characters from newspaper comic strips. His attention-grabbing T-shirt proudly declares 'How-To Geek Is Awesome'. The setting is a snug corner adorned with tech posters and sticky notes on the wall.

Here's DALL-E 3's result.

And here's MidJourney's result.

While the visual output of MidJourmey is aesthetically pleasing, it does not align with our initial request. Hence, DALL-E 3 surpasses it in this aspect. Nevertheless, the image still contains a significant amount of nonsensical text. In my experimentation, I found that DALL-E performs exceptionally well when you explicitly specify all the text within the image or when there is no additional text aside from what you initially requested. However, if the image includes unspecified text, it becomes nonsensical, much like in the case of MidJourney.

​​​​Prompt 3: Setting a Scene

The final test I plan to conduct involves creating a visual representation showcasing the arrangement of key components.

Depicting a cyberpunk cityscape reminiscent of Blade Runner aesthetics, the scene portrays a cyborg woman adorned with luminous eyes and cybernetic limbs positioned on the left, gracefully clutching a gleaming apple. Positioned across from her on the right is a robot vendor, with a weathered exterior, casually smoking a cigar amidst a diverse selection of exotic fruits. The street pulses with energy and movement, as drones soar above and vibrant neon signs illuminate the surroundings.

Here's DALL-E 3's result.

And here are all four attempts by MidJourney.

MidJourney once again showcases its artistic flair, yet disappointingly falls short of fulfilling the specific request outlined in the prompt. Unlike DALL-E 3, which can effortlessly recreate the same image with various styles, MidJourney struggles to consistently reproduce the desired elements and arrangement. Here is the identical image, albeit where I had specifically requested a more surreal and dreamlike style from DALL-E 3.

DALL-E 3 Isn't Perfect

Before you decide to ditch MidJourney for DALL-E 3, there are a few major limitations I ran into when testing DALL-E 3 that you should know about:

ChatGPT refuses to generate images featuring copyrighted characters, whereas MidJourney willingly creates fan art of established characters.

Additionally, ChatGPT does not allow requests for the art style of any artist who is alive, whereas you can still make such requests with MidJourney.

Both platforms have restrictions regarding adult content that is violent or sexual in nature. However, MidJourney offers a straightforward appeals process for false positives, while convincing ChatGPT might be more challenging due to its higher level of sophistication. Although my experience with the tool was limited, it is important to note that both DALL-E 3 and MidJourney frequently receive updates and improvements. Nevertheless, the aforementioned limitations are likely to be the most noticeable concerns for the majority of users.

The Verdict

Choosing a clear winner is challenging, but currently, MidJourney emerges as the ideal option for those seeking expressiveness and artistic finesse in their generated content. On the other hand, DALL-E 3 outshines as the superior tool when it comes to producing consistent artwork tailored precisely to your specifications, especially for illustrations or other professional applications.