Check out our Free Browser Extension that helps you with writing prompts for AI image generators 👇 chromewebstore.google.com/detail/prompt-catalyst/hehieakgdbakdajfpekgmfckplcjmgcf
This video is exactly what I was about to make. It is a good and fair comparison of the two. I am currently using both Midjourney and Dall E 3. Have been using them for a week now and I will say your assessment is accurate. The biggest difference is that Midjourney is amazing at creating realistic faces, beautiful faces, sure. But is weak in terms of listening to your prompt. It's actually quite stupid. For example, my prompt says "a woman holding a spear." And it shows a woman holding swords or some deform weapon that is a combination of different weapons. Dall E 3 would nail that no problem. But sometimes Dalle 3 would mess up the faces, the eyes, or blend different prompt into each other. In summary, Dalle 3 is definitely more accurate in listening to the details of your prompt for sure.
First point goes to DALL.E maybe in terms of understanding context better, but I definitely think the images generated by Midjourney where waaay more detailed and creative.
can u explain? im new and would like to know to which one invest my efforts, like what restrictions/censorship/rules DALLE has and what tools has midjourney not all of them but what are the ones that you can say, yeah midjourney is better. Thanks!
I have been using DALLE for weeks and you can improve the images a little more if you put things like photo-realism or ultra detailed or cinematic style etc. ( in chatgpt plus version )
The PopArt point goes to MidJourney. Just because DallE made everything look like a Lichtenstein litho doesn't qualify it as pop art. It showed limited range in the test subjects you provided. Also, the last one was up for grabs. I'd say there were about 50:50 and no clear winner. Thanks for this video, I've been looking for something like this for a while.
The points given are entirely my opinion, you can ignore them and award your points based on the images and choose a winner. However, if we take human-made pop art as a basis, then DALLE 3 is much closer to this style and follows the prompt better. Also, I didn't get Midjourney's results on the first try, I regenerated the prompt 3-4 times because if it's not photorealism, it creates something completely unrelated. And you can still see those weird lips in the middle, the low-detail lipstick, and random shapes in the background.
Dalle doesn't handle all text well. I've been encountering many spelling errors, repeated words, and mumbo jumbo. However, I'm sure it'll get better! Great vid!
In the vintage photos category, I think Dall-E 3 was the clear winner. Midjourney looked more like modern photography with filters pretending to be vintage. Maybe that's what you want, but that's not what the category was supposed to measure.
I still think Midjourney is a better AI Art generator. I admit I may be biased, but if you want to up your AI art prompt game then Midjourney is better because it uses discord and even though you are using AI, you are still creating in a community amongst other artists and humans through channels and learning new techniques from other humans... It has that perfect balance between AI and human. As an AI generative artist you will learn, explore and improve your AI art prompts way faster with Midjourney than Dall-E3.
"Prompt game" for Midjourney is a myth. I've created over 6000 images with Midjourney, and I can say that it can create visually appealing images, but it completely ignores huge portions of your prompts, no matter how well you describe or phrase them by adding camera model names, artist names, shutter speed, and other unnecessary words alike. You may randomly get an image close to your prompt and think that you are now a “pro at the prompt game”, but this is pure luck and coincidence. Try to re-roll it and you will get completely different results. It only works well with short prompts. For me, an AI image generator should not be a slot machine. Its goal is to take every detail from the user's prompt and put them together in a clear and connected way in the final image. Midjourney only works well with short prompts and photorealistic images and is more like an image compiler than a true AI image generator. Although I agree that the community is great and you can learn and get inspiration from there.
Thank you. Midjourney is great with photorealistic faces, but too creepy-looking because of the machine's way to interpret emotions, or god forbid, hands, lol. I just joined Midjourney and my prompt was "a happy clown holding a paper (with a message). And that clown is straight out from horror movies, as creepy and psychotic as can be. And the simple words are messed up and not legible. Midjourney is a strange mix of amazing, creepy, and inspiring. I'm gettin Dall-e next.
True! Midjourney style is conventional, no surprise it is very good with realistic images. I like DALL E 3 more because it is more experimental, and I get more ideas from it.
Midjourney is good for photorealistic images. Dall-e (which is now integrated into chatgpt) is good for using normal language but then the results are a little bit weird and too stylistic.
Here's why this is not accurate. each of these platforms LEARN based on previous inputs, so your outputs are very dependent on what you've used the platform for previously. for example; if you mainly looked for painterly styles on MJ, when you type in "pixel art" its more likely to skew results toward painterly still.. and Dall-e is the same, and is also dependent on anything chatgpt knows about your interests. So each person's results are gonna be skewed from this
So im curious. Does Dalle learn to create better or more accurate images through people using it and trying to give more detailed prompts or is it just updated by the developers to gain more accuracy?
DALL-E 3 is definitely photorealistic, but by no means realistic. As a matter of fact, this was a deliberate choice. DALL-E 2 was able to generate perfectly realistic photographs...
I thought both were good... the Dalle is more 3D-type drawing, while the MidJourney is more realistic... so for me I like the realism more MidJourney is better.
I see your point, but I think DALLE 3 is more impressive in this regard. After creating over 6000 images with Midjourney, I've come to the conclusion that it can create visually appealing images, but sometimes it completely ignores huge portions of your prompts, no matter how well you describe or phrase them. For me, the main goal of an AI image generator is to take every detail from the user's prompt and put them together in a clear and connected way in the final image. Midjourney only works well with short prompts and photorealistic images, and is more like an image compiler than a true AI image generator. I think v6 will solve this problem, Midjourney already has a great visual component, they just need to make it more responsive to prompts and train it using more images with different art styles.
@@ai.catalyst I understand your point of view. Midjourney has "Midjourney faces". it is easy to generate good pictures with Midjourney, but they all look the same because of the strong AI correction. I always had to try to destroy the "Midjourney face" and create my original facial style. On the other hand, DALLE3 has a strong language understanding, so I think the AI's interpretation is less likely to twist everything. So you are right. It is an interesting comparison because the two AIs have different cores but the two converge. P.S. Thanks for reply! Good discussion. I am not a native English speaker, so please forgive my cheap English.
Dall-E is just cartoonish kids play, no matter how realistic you want, Dall-E can't deliver it, where Midjourney is a matured adult, which provides extremely good pictures. If you ask it it to be realistic pictures, it generates realistic, no fuss.
GUYS GUYS u all are saying DALL E can’t do photo realism and while Midjouney is better, DALL E 3 can do VERY realistic images if you prompt it with like a simple “A dog” or tell chat gpt to prompt DALL E with “A dog” TRUST ME you will like the results. You can also use my custom GPT FutureFusion which can make better DALL E images!
Sorry for the dumb question. But is it possible to keep change details about the image you create? So you can slowly but steady correct the same image further and further?
The video is good. Kindly add one more point which is probably the decider for a lot of stakeholders that AI image generator caters to. And that point is "RESOLUTION". Dall E 3 can only generate images on 1024X1024, whereas Midjourney can generate images at a given resolution. I already had Midjourney, after seeing this video I thought of trying Dall E 3, now I'm stuck with this "RESOLUTION" limitation which pretty much renders all the edge Dalle E had over Midjourney futile.
You are REALLY really biased towards Dall-E :D Even in the comments you reply to all people that disagree trying to argue your point! :D I would have give most points to MidJourney. The pop art by MidJourney was pretty bad though.
DALL E was definitely better than Midjourney when it came out. It's not just about the aesthetics of the images, but also how much of your prompt is actually used to create the image. Midjourney v5 felt more like an advanced search engine. Now that v6 is out and DALL E has been censored into oblivion, they are pretty much on the same level.
@@ai.catalyst And you are doing it again! :D hahahahaha Most people don't give a damn. The image you referenced where Midjourney apparently didn't add something you mentioned, I was so confused about what you even mean. The Midjourney picture looked 5x better, but you gave the point to Dall-E because of a prompt detail I was not even noticing. It's like a woman waiting for her man to do something wrong so she can cheat. 😂😂 "I cheated because he cooked the pasta too long.. the relationship was wonderful but the pasta? That's a no go" Thats the level you are on with this. :D I want images that look good, if there are some details missing, I don't care as long as the overall image is good. The logos looked 10x more professional on midjourney but you gave it a tie? lmao (that's me saying this after having a logo for one of my companies done by Dall-E 3) You are looking for perfection rather than quality.
Award your personal points and feel free to use the image generator you need for your purposes. This is just a comparison and my opinion. My point is that it doesn't have to be a slot machine and you will feel the difference when it is your prompt/request and you don't see the details you need.
Two questions. 1) When I use Dall-E 3 in Chat GPT, it adds text to images but *constantly* makes spelling errors. It's so frustrating. How do you fix that? 2) I understand both take text prompts, but I thought MidJOurney takes photo prompts? Is that not true? Why wasn't that mentioned in your comparison? Thanks!
You can regenerate the images until it gets the text right , or shorten the text if it's too long. If you have Photoshop, you can use generative fill to correct spelling errors. In most cases, Photoshop recreates the font style perfectly and can replace the misspelled letters. Midjourney can take photo prompts but DALLE can't because Midjourney has more features which I mentioned near the end of the video. This comparison was specifically for text prompts.