If you want the full prompt i used to prime GPT-4, sign up for my newsletter here and get a free PDF with whole input text: www.allabtai.com/newsletter/

AI people: "Okay we made artists unemployed, but they can still train in making prompts" Also AI people: "We automated prompts"

What fascinates me the most is how robust GPT 4 is. You don't need to learn some programming language. You can make it learn concepts literally through plain English. The promise of Siri and Alexa is what GPT4 actually is. It's fascinating.

Just wait until gpt-4 gets its image input turned on and then connect midjourney v5 directly on the gpt-4 input. ChatGPT will be able to see what it's making with midjourney. This will get insane 🔥🔥🔥

I really like how you showed the tuning process for GPT4. I think this is crucial and a really cool way to see how one can leverage these new tools to perform on their tasks of interest.

Very cool how you "taught" GPT-4 about Midjourney, and then had it help you generate some prompts. Just amazing!

I didn't see prompting as an issue for artists I saw it as a tool. With this development I see it as a definite issue. The detail and realism is amazing, especially at this early point in the "game". One note" This is not "photography" it's illustrative realism. Photography is the process of capturing light on a receptor through a lens.

2 days ago GPT4, yesterday Midjourney_v5, and today Microsoft 365 Copilot - crazy time to stay updated. Kris, thank you for the video - keep recording...

Very cool to see how putting them both together gets better results could be interesting to put in the same prompt you ask GPT directly in to MidJ to see the difference of a naked prompt vs an ai aided prompt.

It's staggering the uplift in quality when you get GPT-4 to really go to town on the elabroate prompting. It's teaching me to restructure how i write my own prompts, what works, what doesn't .... but MJ5 + GPT4 is a crazy combo Thanks for the PDF!! Really insightful

I wouldn't call this photography. Its just art that looks like photography, but I believe AI is going to hit photography hard, and there are tools to replace a lot of these things and open up different job avenues.

I tried using gpt 4 for Midjourney prompts, and this was before Midjourney v5 released yday and even then I got some really cool images. Now to test the prompts on v5 and compare! Great video

What a time to be alive honestly! I mean, AI is going to be like how the internet went wild in the early 90's! I can't wait to see what comes out in a few more years! It's exciting and sort of scary at the same time!

AI went from slowly crawling to sprinting full speed, and it`s good to see people are on it, finding new ways of improving it`s usability. Awesome stuff.

I tried feeding these steps into the Version 3 and it's the difference is pretty big. It's wild so much better version 4 is at understanding the information getting fed into it. Good video my dude

I’m waiting for the day when they combine midjourney and gpt so that the AI can recognize what the results are and you can simply tell it to make adjustments as you would tell an artist to do. Imagine simply typing “I wanted the sunset to be a bit more orange” and boom it goes ahead and spits out your image with a more orange sunset but without changing anything else.

I was doing this last night! I pasted in parts of the manual that seemed relevant, but you’ve included so much more with the prompt examples and the v5 specific guide. I tried to give gpt prompt samples last year with gpt-3 DaVinci 2 and it was terrible lol. I’m going to try this method. Image prompt writing is so new it wasn’t really trained on the idea yet. Midjourney 5 and GPT4 is incredible.

AI integration is the evolution and future of digital art, photography, commercial advertising, gaming, animation and story telling. AI "editing" in photography is more akin to the evolution from film to digital photography, and the introduction of digital editing software. The food photograph glow-up is a fantastic example, and is what I'd assume is the future of AI in photography looks like, in the immediate future.

Great tutorial on priming MidJourney. I'm looking forward to two updates in MidJourney. 1.) Ability to get hands generated correctly 2.) Ability to generate text / text on graphics

This is awesome. I recently tried Blue Willow, and I'm blown away by the results. I want to know more about it and It would be fantastic if you could make a series tutorial about Blue Willow

I already had a glimpse on what gpt-4 can do if I prime it, but you went to entirely new level! I’ll for sure use your approach. Thanks a lot for sharing!

It's getting closer... but... I've come to believe that a huge issue, currently (bring on v6), is MidJourney trying so hard to treat a single composition in it's totality and tie all composition aspects back to the biases it has "learned". Sure you can dive into weighting... but the next step for MidJourney's evolution could possibly be introducing prompts like - - face, - - left foot, - - right foot, - - torso, - - horns, - -background, - - foreground... so that users can then give MidJourney a little more help deciding what to do and a little more power to treat different parts of the composition independently rather than as a whole... and THEN try to rationalise it all into one composition (or not... if the user really really really WANTS a very disparate composition they should be able to indicate such with a - - frankenstein value (for want of a better term). Obviously this applies mainly to compositions featuring humanoids. Those using MDJ to make websites would want a similar ability to compartmentalise.... but with terms like - - header, - - navbar, - - menu, - - slideshow, - - footer, - - sidebar. It is possibly time MDJ learned how to separate so that, for example, 1 - the horns I requested on my Tiefling don't just look like horn shaped hair (because it was too close to the hair), 2 - the black leather magic tome I requested in his LEFT hand isn't set on fire simply because I also requested a Fireball in the RIGHT hand... and 3 - the little girl can go barefoot. Heck, then maybe even take it further and allow - - cellref C5 where A to K is the distance along the horizontal and 0 to 10 is along the vertical... so I can request - - cellref C2 barefoot - - cellref C8 barefoot.

MJ is really good but my favourite text to image generator is Blue Willow. I produced some incredible artwork with Blue Willow. Blue Willow recently launched V2 with a brand new model upgrade. More importantly it's completely free

Holy crap. That is really good. Not a doubt in my mind that companies would pay a decent amount to have images like these created. The Viking girl screaming and the close up of the bees were both absolutely phenomenal.

You can let create it code :). Then it defines the exact location of the object on the picture. I didnt pay yet so i always run into character limitations but it helps a lot to create pictures that basically all look the same. What also creates amazing images, feed it abstract words "Reality Dysfunction" and stuff like that and put at the end a single defining Object or Name. It will create totally sureal stuff where you still can detect the last mentioned object or name.

Priming chat gpt with prompts before requesting is absolutely genius and the new mid journey works are absolutely stunning

This is amazing. Sorry if this is a stupid question, but is it possible to use these midjourney promts to drive stable diffusion?

This has truly blown me away. I trained my chat-gpt with your prompts and the pictures i'm getting off midjourney now are now next level. Thanks for sharing :-)

This video is by far the most accurate and amazing way to learn how to utilize both AIs to get high quality images back for inspiration and or commercial use. This literally upped my game to another level. Thank god for this video sir, you are a saint 👍🏻👍🏻👍🏻 (quickly generates more images before people watch this vid more and learn how to use it better XD)

In the future, ChatGPT will be customizable, allowing users to interact with the program through speech. In addition, users will be able to choose different "Skins" to personalize the ChatGPT interface. These packages can be purchased at varying prices, ranging from basic secretary models to advanced models like the T800 from Terminator. Those who do not purchase packages will have access to text models that are more advanced than current ones, but still limited in terms of available options. Another detail is that the registration process for access to ChatGPT will be much stricter in order to control all the content that is being created.

I was starting to think I could become a prompt engineer once my current job becomes useless because of AI. Now I find that you can automate prompts as well. Damnit all. Guess I'll be a coal miner in my future.

Would have been nice to compare these images with those you get by crafting some "naive" prompts yourself rather than using GPT4. But pretty neat process, "teaching" ChatGPT things this way seems really powerful.

Thank you for this. I am an elementary STEAM teacher and you helping me, and my students, understand the power of technology, and the appreciation of art.

The evolution of AI from image generator like Bluewillow or Midjorney to giving chat GPT the ability to see is undeniably evident. Actually, it resembles the AI robots that can already recognize photographs and other things. It appears that AI technology will continue to advance

Interesting. But none of these images are difficult to generate using Midjourney alone, and better images can be made by reacting to them on the spot and improving your prompt for the next set of generations. In my experience, a prompt longer than 20 words changes little if anything in the end result.

Thank you for this, have already made great use of it! Could you describe how you go about finding great example prompts for a certain type of image?

As a developer who studied machine learning and has a deep knowledge of how these models work, I'm even more dumbfounded than the common person. I get Midjourney, that's relatively easy, but GPT-4 shouldn't be this good. And yet it is. I'm very curious to see how it pans out.

I would like to see it generate normal looking people - I think it would have trouble because it has trained on media - primarily selected photos of photogenic people.

Dude, this is just brilliant! I never knew about priming the GPT or the prompts for Midjourney, this I have to try myself...very cool stuff!

Great example of priming GPT-4. Have you experienced any problems yet with the length of your conversation? Very long prompts might exceed the maximum sequence length, and for very long conversations, some of the conversation history might fall out of scope, i.e. GPT-4 will not "remember". Have you experienced any of that? Do you have access to the 32k token model, or are you using the default 8k token model?

I have a friend that does food photography for a living, the details in the photo that you prompted with the pasta is on par if not better than what he does. Now imagine midjourney in 5 years, it will probably be able to create movies on his own, you just write the story with him.

now it actually feels like living in the 21st century 👀 amazing work bro, i liked how you integrated gpt4 and midjourney 5.

The issue I'm having with midjourney is when an image is amazing and spot on, but there is slightly small issue with fingers/eyes/unrealistic imperfections. There is no way to tell it "I want this, but fix this area". You need to regenerate and get almost perfect images with different issues. The problem is that if 2% imperfection in the area of the eyes can make the image totally unusable. If there is a way to overcome this issue, I think it will make a great video (it's a common issue and no one posted a solution on youtube yet)

Yes I love the images! Training GPT for Midjourney with currect docs was great. Thanks for the tips!

This is really great especially for "Realistic Photo" images. How would you use your prompt protocol so MidJourney would create images in the style of a well known painter especially if the painter's style is not Photo Realistic?

I’m very experienced with V5. It’s actually detrimental to input such detailed prompts currently. Keep them short and straight to the point.

This is eye opening ! and you explained the process so well. 😄 Can I possibly use BlueWillow in this process?

these text to image ai systems will use gpt4 to filter unusual pictures like weird fingers. Because gpt4 has eyes and detects unusual details. That's gonna be how gpt4 will actually revolutionize ai art

Sadly it only seems to work with GPT 4, because GPT3 doesn´t react to the READ command anymore at some point. Also I think that the prompts you provide in the newsletter pdf contains some errors, not sure how much it affects the result. But maybe I´ll try it in the future if I buy GPT 4.

This is so cool! I guess there soon will be an already trained GPT-4 to do this? So we don't have to train it first?

I'm already exicted about the image feature of gpt-4. Could be really useful for assistants

I tried the prompt and it does get Chat GPT to produce a very detailed Midjourney prompt that certainly SEEMS like it should produce a great result, but it just doesn't. Many times the result looks like a painting, and not even a very realistic painting.

I asked GPT a basic question about a prompt to feed Midjourney and it already added to my very basic description. "For example, you can try “Generate an image of an old brass microscope similar to Leitz, Zeiss and Watson microscopes of the late 1890s”. You can also add more details such as “The microscope should have a brass body with a black japanned base” or “The microscope should have a rotating stage with two stage clips” to make the prompt more specific". I was actually impressed that it was aware of Japanned bases, rotating stages with slide clips, which were common back then.

Great, so we're past the point where you can identify fake images (or videos). I will never trust my eyes again O_o At least AIs can render us new episodes for canceled TV shows ;) Prompt: Season 4 of Lie To Me, please.

What a brilliant idea to leverage ChatGPT to maximize results in Midjourney. I'm in.

The big questions is, is AI/ChatGPT moving towards making us live like house pets? I wonder when Artificial Intelligence started the process and methods of ruling over humans in a futuristic society? We are walking right into it, yet we are fascinated by it.

Did anyone have trouble with the Prompt #5? I got a warning when I used it saying "cutting" was banned. For this prompt: An ultra-realistic, highly detailed portrayal of a futuristic robot... MJ: The word cutting is banned. Circumventing this filter to violate our rules may result in your access being revoked.

Thank you for this video, you're amazing! this really shows the power if AI. Small question: Could you provide the sources to where to use this methods? I believe Stable difussion stopped having a discord server, am I right? I see you typing in a discord server. Thank you!

I feel so lucky to alive to see us come so far since walkmans and bulky new computers to this AI era race. 😊 have been binge watching all content… to be the first to make some real money from this. 🎉

Love Love Lovely!! Wish you give an attempt for GPT-4 + Midjourney V5 for a consistent character generation for story books .. Same character with different poses and at multiple location and with different facial expressions ... Is that possible to achieve now?!

I used your prompts as inspiration to prime ChatGPT to write multiple choice questions for the English Bar Exam, it did a terrific job

I followed your instruction and gave chatGPT the 5th prompt, then it automatically made a 6th prompt on its own. blown away.

This is beautiful now lets figure out how to push something similar to this and create an Unreal Engine 5 workflow/interface GPT3/4>MidJourney>Workflow>UE5 Output - Sounds Crazy but I give it a few years before we can do this

It does great with photo realism but struggles more with abstract concepts

You should tell it to output only the prompt and inside a code block so you can copy it in one click

I still get hung up on the fact that the AI actually understands what you're saying. This is insane.

Great video, thank you for the information! Though I would be hesitant to request the AI to conjure up visions of them ruling us… The day when that becomes a very real possibility is rapidly approaching. Lol

Hi there! 👋 It would be easier to copy and paste if the document was a markdown instead of a pdf. It's hard to copy "Compatibility"; I had to copy it from Midjourney's site. "6" can be multi-line, with each prompt in a new line. , I just wanted to let you know that this has the response from GPT also pasted. You can also move all the prompts under on big field. Thanks for these. By the way 👍

1. Would you explain insertion of a photo into Midjourney and how it creates or modifies the image? Secondly, What if we insert a photo of the last century , (how do we insert and then aske Midjourney to improve the image quality without changing the content or background? Any examples of the second?

Is it possible to upload a picture and without changing it, be able to make look as best as possible? Like improving the light or the contrast etc?

Just as an FYI, this probably wasn’t using the documentation you gave it, only the couple examples. The reason is GPT-4 in ChatGPT has 4k token context window (you can check by asking it about documentation, it just won’t “remember”). It’s doing this without docs, only prompt examples. But GPT-4 API can support up to 32k context window.

still has huge imperfections with body parts, and has that sterile AI feeling to it, and i got some watermarked meatballs so that questions the whole ethical approach of AI generators

I'm completely blown away at the quality! I think I heard you mention that I can use an original photo and then have a I clean it up? Do you know if it's possible to take a background photo and have a i add a different picture or combined two photos. For example a picture of a restaurant without any people and add people sitting at the tables?

thank you for this amazing tutorial! Wow!!! I wonder how you can get good results when using a real photo of a person and you have to combine it with a stye, because when I tried to modify myself, it doesn´t look like me in most cases. Or if I want to place a product picture or render into a generated ai scene. It always get´s chaotic. Any solutions for that?

I would love to see a comparison of inputting the gpt prompts to the simple original prompts

Well, this is way better than most of human work, absolutely impressive

I’m curious about the need to prime ChatGPT. If you primed it with those learning inputs, does that not stick? Why would I have to do the same thing? Is it’s knowledge associated somehow with me?

Even food photography is not going away! Because if I need to do a new shot of a dish for a menu, I would still require a photographer who can take a photos of dishes, instead of generating some random food. Same goes with wedding photography or wild nature photography, or even the green comet (which recently was shoot on cameras all over the world) ☄️ Maybe for a mock up of a website it will work if you need a photos of random people or random food, instead of using unsplash like I’m doing now. 😅

The problem with Chat GPT is trying to think of things to ask it to do so this video was very instructive. thanks for taking the tim eto study these pics. Where will this go?!

That was a great video! I have looked at midjourney but this really made me sit down a give it a shot. well done!

The biggest problem is that GPT4 STILL has a very short memory and starts forgetting past context very quickly.

I got to try this. These results look absolutely stunning. Wow

Hi! Thanks for the video and your hard work. Question, didn't you exceed the context length tokens allowed? Do you make some test to really be sure that the AI is taking into account everything you paste in the chat window?

One my concern with making high skill level art easy through AI is that it gonna turn it generic and not interesting as a result of there being too much of it and it will just feel like soulless stock photo compilation

This is working incredible, you can also teach it to write the prompts in a code block, so you can copy it faster!

Hmm i did something similar with GPT-3, but i had to make it create a story character, explain what format to use, then have it tell me what prompt that character would give to DALL-E to generate an image.

WHile the results are easy and amazing... as an artist I feel this eventually puts me, and millions of others, out of work.

I'd like to see you put in "dumb prompts" like just going to Midjourney and ask it to draw a plate of pasta to compare the two. Amazing stuff!

This is very important as not many people have the knowledge to be this detailed in explaining the prompts. Soon it wil work without priming.

I almost feel sorry for ChatGPT that it cannot view its own good work. And also learn from it. For me, I like to use Midjourney and make adjustments and try again. Yet ChatGPT doesn't have that capability.

So essentially they are close with replacing all photoshop/ photography creatives , it's kind of sad for people who trained themselves for years in in this

This is amazing! Thanks Kris for the Tutorial. Extremely helpful 💌

Actually, the realistic result was ALREADY available in V3, people just dont know how to write prompts properly to render a very realistic one.. Now.. the only problem with v3 or v4 are the hands. Now, in v5, they somehow fixed it. I hope it stay that way.

Great video. And great ideas for the generating these great MJ prompts . However I have tried signing up to newsletter and the PDF, but sign up doesn’t seem to be working (now tried with two different emails)

We are so close to making everything obsolete. The interior design thing for example. You let the computer design your interior, then it sends the information to 3D printers to actually produce the stuff it designed, their computers send you a quote, which you pay online, then a robot loads the 3D printers with the needed printing materials, a packing robot loads it into an automated truck, which delivers it to your doorstep where you've hired a couple carrying robots who put the stuff exactly into the room as it was planned. The only time a human is used in this whole process is when you type the prompt to get an interior design for your room.

This is truly mindblowing! I've just started dipping my toes into this and it's both overwhelming and absolutely fascinating!

It's not photography, it's still digital art, still amazing tho :)

Just stumbled on this clip - amazing video mate and the best channel around on exploring these great AI engines

It seems to me you exceeded the token limit that the ChatGPT integrated version of GPT-4 would have. So a lot of the things from the beginning may not have been retained in detail. You probably would have gotten similar results only posting the last few posts

I've wondered if the "camera" descriptions actually do anything for the image aside from put a camera of similar make into the image, or is that the trade off. That yes sometimes it gets interpreted literally and therefore we see a "camera" but we also benefit in other images from what a camera does when "taking" a picture with those descriptors. Anyway your the reason 😉 why Chat GPT-4 has gone from 500 messages every 4 hours to 25 messages every 4 hours. Keep up the creative work. 😊