OpenAI’s latest version of DALL·E, a cutting-edge AI image generator, is now part of ChatGPT Plus.
Having used both Midjourney and ChatGPT Plus, I’m now considering whether the addition of DALL·E to ChatGPT provides an opportunity to cancel my Midjourney monthly subscription.
When it comes to transforming text prompts into compelling visuals, both DALL·E 3 and Midjourney do a good job. But which is the better tool for crafting standout images for blog posts? I completed a series of tests with both to discover the answer.
In this post all the examples start with the DALL·E 3 image and the Midjourney second.
Creativity
The most noticeable difference between the two is that DALL·E 3 is significantly better at understanding user intent. It’s adept at deciphering complex prompts and normally produces images that align closely with your vision or the topic you want to illustrate.
For instance, if I’m uncertain about which image to use for a blog, I could ask ChatGPT to suggest a suitable image. Doing the same with Midjourney often leads to unusual result
In a test, I prompted both tools to:
Create an image for a blog titled ‘Domain names you should register to protect your brand’.
The result from DALL·E 3 was closer to the expected outcome. Although not perfect (for instance, the text on a shield in the image read ‘Domnis’ instead of ‘Domains’), DALL·E 3 allows for corrections, which it often gets right. In contrast, Midjourney produced an image totally unrelated to the request.
Realism
Both Midjourney and DALL·E 3 manage to produce realistic images of inanimate objects. Depending on how large you use the image, it’s easy to be convinced it’s genuine.
Both tools generated decent results from the prompt:
Bedroom from the 1930s
However, Midjourney excels when it comes to images of real people or animals. I’ve struggled to get DALL·E 3 to produce believable images in this category. For instance, when prompted with:
Labrador looking happy, animal portraiture
The results for DALL·E 3 weren’t as lifelike as I’d hoped. Midjourney, although obviously an AI image, is a lot closer to looking realistic.
Licence
According to OpenAI terms, you own the images you create with DALL·E 3, regardless of whether you use the free version or pay via ChatGPT. This means you can reprint, sell, and merchandise these images.
In contrast, with Midjourney, your rights to the assets you create depend on your subscription. Paid members own the assets, while those using the free service do not.
Image aspect ratio and size
Currently, DALL·E 3 can produce images in three sizes:
- square: 1024×1024
- tall (portrait): 1024×1792
- wide (landscape): 1792×1024
Although you can choose between square, tall, or wide, DALL·E 3 doesn’t allow for specific dimensions.
Midjourney offers more flexibility in this regard. Users can specify ratios tailored to their needs. Their default sizes are:
- initial size: 512×512 pixels
- upscale size: 1024×1024 pixels
Overall
In general, I’ve found DALL·E 3 more accurate in creating what I’ve asked for, making it a less frustrating experience compared to Midjourney.
Both tools can generate compelling blog images. Midjourney has a slight edge in realism, especially with images of people and animals. However, overall, I believe DALL·E 3, integrated with ChatGPT, offers enough value to consider ending my Midjourney subscription.
5 January, 2024 at 6:36 am
Awesome comparison article and I agree with your points of differences. I think Dalle is way better at understanding what I want and the astounding ability to understand natural language make it way easier to use than Midjourney.
I’m also learning Leonardo.ai, mainly because I watched a YouTube video and read some articles on how it is the best at creating consistent characters. Do you have an opinion on which AI does the best job of this?
27 January, 2024 at 4:10 pm
As a professional artist, non-digital, I lik the fact my creations will be Copyrighted.