DALL·E 3 vs MidJourney: Which is better for business
Which generative AI art system should you use?
Hey — It’s Hussein 👋
San Diego finally has a winning sports team!
Last weekend, we attended the last regular season game of the San Diego Wave FC (our women’s soccer team) and saw them finish the season in first place and take home the shield (aka trophy)! ⚽️
San Diego Wave FC Crowned Champions
If you haven’t been to a women’s professional sporting event, I highly recommend you give it a shot. The game is just as intense, exciting, and fun to watch while tickets are cheaper, crowds are smaller, and there tend to be fewer fights and drunks yelling.
Today’s topic has been on my list for a while: comparing DALL·E 3 vs. MidJourney with a focus on which is better for businesses and startups to use.
Before we begin… a big thank you to:
What are DALL·E 3 and Midjourney?
Both are generative AI art creators. You give them a prompt, they create an image.
DALL·E 3 was created by OpenAI (makers of ChatGPT) and is now integrated within ChatGPT (if you pay for Plus). The older version DALL·E 2 is still available outside of ChatGPT and gives you free credit to use if you want to try it out. In this article, the focus is on features available in DALL·E 3.
To access DALL·E 3, go to ChatGPT (web or mobile app) and check the GPT-4 menu. You should see DALL·E 3 as an option. Choose that to get started.
Access DALL·E 3 in ChatGPT
An independent San Francisco company developed Midjourney, which is only accessible on Discord. If you are unfamiliar with them or Discord, read my intro to Midjourney guide.
To access Midjourney, go to Midjourney.com and create an account. You’ll also need to create a Discord account and join the Midjourney server on Discord or create your own.
For help, see the article I linked above or hit reply. Happy to help.
How businesses can use Midjourney and DALL·E 3
I rely on generative AI art to create logos, stock imagery, illustrations, feature images, etc. I also sometimes use them to have some fun with my kids.
Instead of listing 10 different ways you can use it in your business. Here is a prompt you can use in ChatGPT to help you come up with ideas for your industry and your specific use case that would be more relevant to you:
✨ Pro Tip: you can simplify this prompt by setting up your ChatGPT Custom Instructions which eliminates the need to explain who you are and what your business does.
Differences in prompting
DALL·E 3 is integrated within ChatGPT, which means you can use ChatGPT to write an image prompt for you, while in Midjourney, you have to create the prompt yourself.
For example, let’s say you are a food blogger and wrote an article about various sauces that go well on steak. You want to share this article on social but need an image to go with it.
In ChatGPT, your prompt could look like this:
ChatGPT will then turn your request into an image prompt and pass it through DALL·E 3 to create the images. It will also give you the prompts used to create those images, which you can reuse in the future.
What was great with the output from DALL·E 3 is that it created a collage of four different images with different steak sauces without additional prompting. It also gave me four different options of the collages to choose from. All of this took seconds to create.
Meanwhile, in Midjourney, you have to come up with the prompt yourself, which would look more like this:
Both images are perfectly usable for our use case. You might say that Midjourney’s photo looks more realistic, but I had to spend a good 20 minutes to end up with this photo, while with ChatGPT, I got the image on the first try. I could go back and ask it to make it more realistic and make some modifications if needed.
Regardless, with DALL·E 3, prompting is much simpler.
Generative AI art and copyright
Before you read this section, keep in mind that I am not an attorney. If you are concerned about the copyright of your generative AI work, please consult with an attorney.
From the perspective of OpenAI and Midjourney, you own the copyright of the work you create with their tools. However, the law may not agree.
A federal judge recently ruled that generative AI art cannot be copyrighted because it lacks human authorship. That could be debated since a human had to create the prompt to create the image. I’ll keep an eye on this issue and keep you up to speed as the laws unfold.
OpenAI says that images you create with DALL·E are yours to own and do what you like with them, as long as you follow their terms and conditions and do not deceive people into thinking AI was not involved. I didn’t see any mention of them wanting a license to the images you create. (source 1, source 2)
Midjourney’s policy is a bit more complex. Depending on the size of the company you work for, you need to be subscribed to a specific plan in order for you to own the copyright of the images you created. You also give them a license to use the images as they see fit. (source)
Additional differences and features
Text in images
Neither of the two generative AI Art products works well with having text in the image. However, DALL·E 3 is much stronger with this and will usually include words that make sense. Midjourney text is always non-sense.
Describing an image
With MidJourney, you can upload an image and ask it to describe it for you. That will result in four prompts that you can use to generate similar images.
With DALL·E 3 this does not exist, however you can upload it to a new ChatGPT conversation, ask for a description, then go back to DALL·E 3 to create similar images based on the description.
Zoom out (outpainting)
With Midjourney, you can zoom out an image which gives you more context around the image you created. It’s a really fun feature that can be used to create GIFs and videos like the one below.
You can ask DALL·E 3 for a zoom-out of an image it created, which works, but it doesn’t retain the integrity of the original image and often makes changes to it. If your goal is to create an animation, this is not usable. If you want a wider perspective, then it works.
Midjourney zoom out
Image size and aspect ratio
The default image size in DALL·E 3 is 1792 × 1024. In Midjourney, the default is 1024 × 1024. You can ask ChatGPT for a specific aspect ratio in your prompt, but it sometimes gets missed.
Midjourney has a specific function for this, which always works.
Additional features available in Midjourney and DALL·E 2 (not 3):
You can pan out in any direction (extend the picture)
you can upload an image to generate similar images to
You can regenerate a section of a generated image (with DALL·E 2 you can also delete text with this feature, which is useful)
You can regenerate similar images with slight or large modifications
I would expect that these will be available in DALL·E 3 in the near future.
Additional features only available in Midjourney:
Merge two images together
DALL·E 3 requires a ChatGPT Plus subscription. This will cost you $20/month, which is well worth it if you use ChatGPT for work. For the same $20, you get access to GPT-4, Browsing, Advanced Data Analysis, Plugins, and DALL·E 3. You also don’t have as many availability issues as on the free plan.
Midjourney has various plans starting at $10/month and going up to $120/month with limits on Fast GPU time, which you can buy more of when you run out.
Which should you use?
If you are already using ChatGPT, you should start with DALL·E 3. No need to introduce a second tool to your workflow.
The only time I would recommend using Midjourney in addition to DALL·E 3 is if you have a specific use case that requires some of the features I mentioned above (like creating animated GIFs) or perhaps you have specific image quality needs or style that DALL·E 3 is struggling to create and you want to see if Midjourney can do better.
For me, I am using both. I prefer Midjourney for icons and logos and DALL·E 3 for everything else.
But in my case, I have to pay for them both and know how they work regardless since I write about them on GPT Hacks.
There you have it! I hope you found this useful.
I would love to share your best generative AI art with GPT Hacks readers. Post an image on social with the tag #GPTHacks, and I’ll reshare it.
See you next week — Hussein ✌️
P.S. If you’d like to sponsor, reply. 5.6k founders and entrepreneurs are waiting for you.
P.P.S If you live in or visit San Diego, sign up here to be notified of tech and founder mixers I organize in the area. Would love to meet you there.
How did you like today's newsletter?