Midjourney V6 VS DALL•E 3: Prompt Battle & Full Review

Does mid journey provide any apis? Like dalle3?

Bing image creator gives you 4 photos too and is dalle-3

To sum up:
MJ makes photos
Dall-e makes paitings

Dalle is very good at specifics. Heres a free secret, I use this strategy to combine chat gpt and DALLE-3's power to be much more useful then a standard approach. These are some lines directly out of a GPT I made to simulate advanced martial arts positions that DALLE is not trained to create, these positions are far to specific but through this method I get pretty insane results.

But of course first chat gpt knows its job and its expertise THEN i force it to ALWAYS follow a 6 step process. I also included A PDF and 20 images of reference to help it.

I would say something like: Provide of a left hook to the body from a southpaw fighter, its opponent is orthodox. Then my gpt follows these steps:

1. Analysis of the scenario: Thoroughly breakdown the technique/scenario/strategy providing expert analysis that includes the body mechanics explained on a scientific level, using anatomy and biomechanics provide an in depth explanation with your thought processes included.

2. Reference Material Analysis: Further refine your understanding of the any position by analyzing additional reference materials from your knowledge base. This involves a deeper look into the positioning of limbs in combat sports to ensure a more accurate depiction in the graphics. LET THE USER KNOW YOUR ARE PERFORMING THIS STEP BUT DO NOT EXPLAIN IT.

3. Layered Construction Concept: While you can't manually manipulate 3D models, you can conceptualize the images in layers, starting with a basic outline that focuses on accurate limb positioning before adding details. This method can help visualize the correct placement of limbs in advanced scenarios. LET THE USER KNOW YOUR ARE PERFORMING THIS STEP BUT DO NOT EXPLAIN IT.

4. Feedback Loop Adaptation: Although you cannot receive live feedback, you can internally review the generated images against the established principles of hand to hand combat and biomechanics, comparing them with the detailed descriptions and visuals within my training data. This internal review process aims to identify and correct discrepancies related to limb positioning. LET THE USER KNOW YOUR ARE PERFORMING THIS STEP BUT DO NOT EXPLAIN IT.

5. Enhanced Description for Image Generation: I can refine and elaborate on the description used for generating images, emphasizing the accuracy of limb positioning and the anatomical correctness of the fighters' stances and movements. This involves being more specific about how limbs should be positioned in relation to each other to avoid issues like limb dysmorphia. (this is generating a reusable and adaptable prompt)

The last step is simply to generate the image.

Love this channel, hope someone finds this little gem on prompting Dalle helpful.

However you can get either image - and both types of image with the proper prompt - you would simply edit one prompt to make it more comparable to the other - we are just looking at an interpretation, I would follow up with , worn and rust but full of heart, highly textured surroundings and so on..

the dragon by Dalle looks so much like a John Garland painting that I have a feeling that it could contain huge parts of this works...

Yeah I stay off of discord,too many negatives for me ,too many stories of people getting hacked and or stalked 😊

Both are for no talent hacks

please review your Alphonse Mucha pronunciation, its Mu - ch -a : ch as "Koch" the ch without k

The issue with the fantasy dragon image is that there are things that don't make sense. The castle in the foreground makes it seem that the dragon behind it is larger than a building. Also the valley is lit by sunshine but there is a night sky. MJ wins that one hands down IMHO.

thanks for this detailed comparison.
To me, your comparison revealed that DallE was mainly trained with Disney-content, and I mean the stuff for kids. While Midjourney was trained with all kinds of stuff from photos to art paintings to (maybe) Japanese anime. Disney content isn't a bad thing, they really have high quality standards, but the kids stuff is too polished, too composed and it always carries this forced happy mood where you expect someone starts to sing and dance every second. Emotionally, it's one dimensional. Midjourney seems to be better in carrying all kinds of moods, not only the happy ones.
But now to the downside of BOTH. They are censored. And not only in a way to prevent illegal stuff, but to be child-friendly. PG10 I would guess. The problem is, most paying customers are adults. For software that is created and maintained in the free part of the world, it's hard to believe how strong they enforce censorship in their products. How hard would it be for them to offer two different models, one child-friendly and one that's almost uncensored for people who are old enough to decide for themselves what they want to see and create?

Nice content man, very precise. Thank you

The difference is Midjourney is doing a lot more hand holding. Dalle's 3D capabilities for are exceptional if you know how to use it beyond a simple prompt.

DALLE can’t create photorealistic images.

I agree I don’t like DALL-E

Midjourney has looked way better since the start

yeah, DALL-E is corny every time. It used to be okay when you could use artists names in the prompts, like cinematic in the style of David Lynch. Now it won't take prompts like that.

31 mins, you gotta be kiddng

Did seemed to me that Midjourney was going to get the palm leaf whatever it did, I am left questioning how neutral this appraisal was. In the robot example, for example, Midjourney completely ignored most of the brief - something it is very prone to doing. Just like a real graphic artist then ...

Great video thanks. I’m doing your children’s picture book course too, where you use Midjourney (when it was free) for that. Seeing as now I’ll have to pay to use AI art, would you say to pay for Midjourney or Dall-e? It seems as if Dall-e might be better for children’s art anyway in this case? Especially if I need certain characters to interact.

DALL-E is cartoonish and painty, it is hard to extract photorealistic images from it, but it is free.
All this txt to image pricing doesn't make any sense. For 10$ you get nothing, and for 30$ you get almost nothing. In comparison, cloud gaming like xbox game pass which grants UNLIMITED server time on same or more powerfull hardware, cost 17$ per month.

As a Dall•e 3 user it’s all about you prompt, comparing the two is not really practical when they both use different ways of interpreting a prompt. You can indeed get a just as good as Mid Journey if you “know” how to prompt, using bokeh, distance etc etc I can indeed recreate just as good, and did with the cyborg image by prompting correctly. It’s just maybe Mid Journey is easier. And yes you can use certain images in Dall•e 3, like Star Wars, Marvel, and painters ‘in the style of…’

this is really helpful contents for making a decision. Great Job! Thank you very much!!!

Please add pixel art to the comparison

Best comparison video ever!

Let's not forget that you can use Dalle for free via the Bing Image Generator, for up to 60 images daily, not counting those eaten by their puritan dog.

A useful comparison, efforts appreciated!

Dalle just doesn't understand framing or composition while midjorney does.

I like MJ much more.

I wouldnt call this video bias at all. He did the prompts and he called out what the results were. MJ results came out alot more realistic whereas the other is more for fun.

When you spoke about pricing, you said Dall e was $20 per month with no restrictions on the amount of images, but do you not also have to pay for each image you generate, as well as the $20?

So glad I went down the MJ route instead of Dalle. Not missing a thing!

You should have compared faces more, dall e always creates the same face...

In depicting dragons, the Japanese would choose the image of Midjourney since symmetry in art for the Japanese...

DALL-E cant even make a man without facial hair. Try asking GPT to create an image of a man without beard or facial hair and see. So horrible

I appreciate this video! Been a Stylar user and I think its AI is a bit more up to snuff. Ever thought of giving it a go?

very useful, thanks!

MJ, DE3, and Ideogram are probably the best models out there right now. After getting really into mid journey 5 and 6 I am pretty much just done with the stable diffusion. I just continue to support the community by putting LoRA from Civitai to generator sites but that's about it. Thanks.

Now, do the same for Israeli people?

Very good all encompassing review. Love the USA guy..what a crack up.

As long as censorship exists they both suck. Get Stable Diffusion

Hello, there are a good number of new videos on these best-known AI art generators almost on a weekly basis, which is occasionally overwhelming. While most of the video content of the other guys are filled with some humorous elements in an attempt to be engaging, there is always this element of depth and richness in your videos which makes the videos more informative and enriching. I also appreciate your quite fine sense of art when you comment on the images generated by the competitive AI generators -and I can't help but smile sympathetically when you incline in favor of Midjourney in regards to its peculiar approach (or philosophy as you define) to the image generation. If I'm not mistaken, you aim to reach out a certain group of people who do take the desire for professionalism seriously and strive for enriching themselves new and genuine information in the field. If this is true, I could tell you that you're exceptionally informative and enriching for them. I personally thank you for the good quality of content in your videos. Greetings (and sincere sense of thankfulness) from Turkey.

I feel like one thing you don't really seem to take into account are the many errors I can spot in MJ's images:

1. The photorealistic image has a black blob on the water.
2. In the illustrated fantasy one the dragon seems out of proportion and its tail wrong and there's also a random wing at the bottom of the image that makes no sense.
3. The tattoo design one features some chicken-type legs but they don't really make sense.
4. In the animal design one, the street lamp is way too short.
5. In the floating islands one the foaming water doesn't look right in the bottom left corner. (Also, not really an error but since you've mentioned repetition in the other image, I feel the moons look repetitive here as well. Just to be fair.)
6. The hand looks out of proportion - esp. the forearm which is too small.
7. For the art novo one, it's trying to cover the breasts of the woman in a way that makes her look like she has none.
8. For the interior one, the ceiling fan is messed up.

Love it

Thank you for the great comparison!
The video is now 9 months old. Has one of the AI's improved more than the other during that time?

Some times MJ is like Dostoevsky.

Midjourney is not good if you plan to make stories.
If you plan on making stories with images then Dall E is the best of the best.
Midjourney is for a different purpose because it's designed for realistic photography.
Adittionally DALL-E can create images of remote and rare cultures which is something MidJ and leonardo arent fully capable of without an image prompt. Dall E. is extremely intelligent because it can understand very complex requests.