Home Creators Posts Import Register Favorites Logout

Content

Dall-E prompt: photo of a male librarian holding a sign let's test dall-e 3

Alas, poor Midjourney

As you probably know, OpenAI announced that Dall-E would be integrated with ChatGPT, opening new possibilities for image generation. A few AInfluenceres (this is how I call them now) were declaring that Midjourney was DEAD!

Wow! Is that right?? Or it's only a way to get more clicks and new followers on X-formerly-Twitter, and receive a better payment at the end of the month?

Midjourney prompt: elon musk in the middle of a bank getting some money and give them to small birds, view from above, in the style of 3D art, octane rendering, ultra-realistic

Dall-E prompt: elon musk in the middle of a bank getting some money and give them to small birds, view from above, in the style of 3D art, octane rendering, ultra-realistic

Of course, Dall-E has a few safeguards and refuses inquiries that specifically request pictures of a well-known individual or in the style of an artist still alive. Personally, I think this is the future of MJ too.  With MJ getting bigger and bigger, the risk of a lawsuit is more concrete. I have noticed that v. 4 and even v.3 were better with some celebrities' names than v. 5, but maybe it was not a deliberate decision yet.

Anyway, let's see what this robot can do! The left image after each prompt is MJ's generation and the image at right is produced by Dall-E. I chose the picture that followed closely the prompt in each bunch and I didn't re-roll, so no second chances!!

Prompt: Princess Peach by Martin Parr

While MJ generated what I was expecting, D3 was a bit literalistic.

My fault.. let's be a bit more clear with this prompt: "A photography of Princess Peach in the style of Martin Parr." D3 finally brought me a person, but not the style I was expecting.

Prompt: futuristic scene with skyscrapers, hovercrafts and robots

Let's try a prompt suggested by Dall-E itself. The result was scenes with some similarities. Personally, I prefer MJ's color but D3 didn't ignore the robots.

Prompt: an Infograph describing the different kinds of pottery

Personally, I think Dall-E brought me a better style, something you would see in a cool newspaper. MJ might require some prompting to bring something similar.

Prompt: Latin-American president, official photography, lightroom, fujifilm, portra 200, 8k, 16k, high quality, insane quality, intricate details, insane detail, high detail, fine detail, exquisite details, high resolution, insane details, extreme details

This is a prompt I tried when v.4 was out. Both produced realistic images, but MJ generated a really detailed picture, where you can see even the skin texture of the subject. His clothes and medals are very detailed too (maybe too much detailed... bias?). Dall-E's production is a bit more "plastic," and includes American (not Latin-American) symbols. (Btw, the other option had a bigger US flag...).

Prompt: Nazgul over sleeping town by Tyrus Wong

This was a prompt for our  April's Fool post. On that occasion, Dall-E reproduced Wong's style (sorta). Now we got a generic fantasy picture (but I liked the guy sleeping outdoors). MJ did a better job following the style instructions.

Prompt: train riding into the sunset, lake and mountains, fantasy acrylic

Another prompt suggested by Dall-E. Both generated good images. MJ is a bit more detailed, and Dall-E showed the tendency to do this composition with the train out of the frame.

Prompt: A cactus wearing a Hawaiian shirt and playing a ukulele

Another prompt suggested by Dall-E. I liked both, but MJ never produced a cactus, only creatures inspired by cactus (at most). Dall-E generated this little cute ukulele player.

Of course, MJ has some aces up its sleeve, like --style raw:

Prompt: A closeup portrait of a playful maid, undercut hair, apron, pronounced feminine feature, kitchen, ginger , freckles, flirting with camera

This is part of a famous Stable Diffusion prompt. It is not a NSFW image, but I had to change some words until MJ and D3 accepted it. (By the way, Dall-E censored more than Midjourney). Midjourney included the kitchen as background, while Dall-E focused more on human features. (and got the hairstyle right).

Conclusion

My idea here was not to foretell which system would prevail in the end. It is clear that MJ and Dall-E have their own weaknesses and a lot of strengths. (Also, both generate good hands!) Personally, I still prefer Midjourney's style and quality, but Dall-E advancements are undeniable.

And how MJ is doing with words? Well...

MJ prompt: photo of a male librarian holding a sign let's test dall-e 3

Tell me what you think in the comments or on our discord!!

See Ya Later!

Jos

Files

Comments

Derek Redican

If you work in advertising like I do, DALL-E produces art director concept-friendly solutions to prompts. MidJourney tends to go 'arty' at the expense of true-to-concept. However, Midjourney is still superior on the detail and resolution fronts.

Gaffer Wiles

I am guessing that the images here *might* have been created with the https://bing.com/create "hack". I seriously question that the Dall-E 3 engine that Bing is using is really the full release of new generator. It might be,but it certainly looks cut-down and "nerfed" compared to what is coming out of the ChatGPT 4 integration that some lucky people have access to.

Gaffer Wiles

You can contact me via email (maybe even Discord chat. ) Our revered MidLibrarian, Andre has my details...