Dall-E 3 with Chat GPT Plus: No Prompting (Workflow Tutorial)

hi I am Mali and welcome to the channel I just got access to Del 3 and I was utterly Blown Away by how they integrated DOI with chat GPT have a look at these images it is not about the image Aesthetics it is more about how you can generate them easily and quickly I did not use a single prompt to generate these images the way we are used to prompting in generative AI has wholly changed let me show you the correct way to get accurate results using Di and chat [Music] GPT before I show you some workflow techniques let me briefly explain what happened here prompt engineering as we know it is dead when you mix an AI model trained with linguist itic texts and a generative AI model trained with images you get a match made in heaven and that's precisely what Del 3 and chat GPT are about you no longer need to come up with a prompt I have tested this extensively on various concept arts and you get exactly what you are asking for every time if you are a chat GPT plus user you will get the D 3 option which is still in beta click on it to enable image Generations let's try something simple say logo for a restaurant here I am not looking for anything specific although you can be I am just trying and testing what the AI can come up with to bounce some ideas notice how I just wrote COC Burgers I did not mention that I want a crocodile eating a burger as a logo chat GPT reads the images in the following order once you click a specific image it will give you the prompt it generated chat GPT understood what I meant by crock Burgers it accurately wrote a prompt for me now let's say you want to build on this image you can copy The Prompt or type in image three and whatever changes you want for that specific image I am just asking it to change the text to Hungry Burgers it doesn't give me the same image because what I asked it to change was a significant element of the image I will show you some examples where it changes only specific elements maintaining the overall image this is dependent on what you are asking it to to change I wanted to see how well it can manipulate text unfortunately this is the limit it cannot distinguish and color each text differently what it did here was merge the text so let's revert it back to Hungry Burgers but in red and it got the text partially right I can take this image and put it in Photoshop for further editing this was a simple test and it shows you how intuitive the AI can be let's try something complicated I purposely type it this way as I don't want to use a prompt I have had a lot of trouble getting the exact color and design when doing some fashion work in mid Journey I want a Victorian style dress with some fractal floral patterns on it notice how I have separated the pattern color and the dress color you can be extremely specific here what I have written is still very generic I did not mention the aspect ratio you can generate an image in square tall or wide aspect ratios in a fashion design I need a tall as aspect ratio the AI just understood that let's focus on the Four Points the AI understood it has given me four image Styles photo illustration oil painting and a render I am specifically looking for something given to me in the fourth image I did not know I had to use the word render in the prompt to generate this style I will try to direct the AI to Showcase a single mannequin instead of three and build up from there check out the prompt it generated it accurately took the colors for the individual elements and applied them correctly try doing this in mid journey and you will go round and round in regenerations mid Journey will give you highly aesthetic images with no accuracy or control I can get this accuracy in stable diffusion however that's still a technical challenge with the hardware requirements and finding the suitable checkpoint trained model now I will tell the AI to isolate the dress into a single dress instead of three again I am just using simple language and no commands or prompts now that it has understood what I want I can make changes to it before I move on to the next sample workflow check out this trick I asked it to make a simple linen shirt with short sleeves and a top pocket and it gives me this simple shirt I am asking it to change the color to light blue and it does precisely that then I asked it to make the sleeves longer finally I wanted to remove the top pocket and look at the level of control you are getting with a simple chat I have no idea why this hasn't taken off but being able to tell the AI in such a straightforward manner you seriously have to give credit to what they have done here It ultimately accelerates your workflow for jewelry design let's go for an engagement ring with a delicate Italian design 22 karat gold and a 1 karat rose cut solitire diamond these are pretty generic again it gives me two photos an illustration and a render let's redesign the ring with a honeycomb Holo pattern in the first photo beautiful and see how it maintains the same perspective even in a photo let's challenge it and see if it can adopt this same design in a bracelet and it does a beautiful job see how the first bracelet design is so similar to the ring design in my opinion even the illustration and the rendered images are excellent let's try rendering a 3D elephant surfing in the ocean okay these are cute the first one is like an actual 3D render however the second and the third ones are what I like another thing that you can do is change the weather and the time so if I ask it to change the time of the day to evening with a sunset it will exactly do that and you can Define it more precisely it does work well as I said before it will use whatever it finds best if you don't mention the aspect ratio and if you don't like the wide aspect ratio here you can ask it to change the aspect ratio to square this can be a bit glitchy when asking it to switch from a wide to a tall one most of the time it just rotated the image to a 90° angle but going from a wide to a square post generation or from a square to a tall is not an issue in one of the previous mid Journey videos I talked about how to use math in mid Journey prompting I learned most of the math formulas and what it does precisely via chat GP therefore I wanted to try out an interior design by just telling it the math formula which I know has quite a few subsets surprisingly it used four different subsets and generated the images for me I gave it a general idea that I wanted an interior design concept for a spa reception along with a tropical location like an island I asked it to use the Fibonacci sequence for the interior Decor I also mentioned the atmosphere should give a calm and tropical vibe let's focus on the prompt to understand better the Fibonacci math subset it used in the first image I can see the golden ratio literally designed on the floor and it made the desk curved and aligned as per the Fibonacci sequence here it used seashell mural art on the wall a classic Fibonacci spiral design however the seating arrangement needs some improvements I don't like this one at all I do see the spiral design illusion from the stairs to the ceiling but I don't see it being practical and this last one is supposed to have the sequence subtly integrated into the fabric patterns unfortunately I cannot see the details now I want to design an actual spa massage room interior based on the second image and this is precisely where DOI excels over mid Journey so you can see how quickly I got these and the second image has the same seashell mural art but on a smaller scale it cannot do humans as realistic as mid Journey but it can do photography pretty well including editing you can ask it to change the coloring of the Aurora Borealis and I actually don't need Photoshop to do this so it does change the terrain slightly but it's pretty much consistent more like a subtle variation another example it generated an image of a cabin I just asked it to change the cabin to a lighthouse and it regenerated this image they are not precisely the same but the variation is very subtle and it's kind of okay for a generative AI to do this you can also use this to create coloring book images game assets graphic comics and much more for example I managed to get some concept designs for a perfume bottle print ad the conclusion is that dii is still in beta and hopefully it will just get better it also saves time in your workflow because it is accurate and understanding you without requiring complex prompting the number of regenerations is drastically reduced as well until next [Music] [Music] time

Channel: Control+Alt+AI

Views: 2,596

Keywords: dalle 3, dall-e 3, ai art generator, dall e 3, chatgpt dalle3, open ai, dall e 3 ai, dall-e 3 demo, openai, chatgpt dall e 3, dall e 3 tutorial, dall-e 3 tutorial, dall e, dall e 3 chatgpt, dalle 3 ai, openai dall e 3, dall-e 3 first impressions, chatgpt dalle 3, dall-e 3 chatgpt, how to use dall e 3, dalle3, dall-e 3 ai, chatgpt plus, how to use dall.e 3, dall-e 3 prompts, dall e 3 access, get dall e 3, get dalle 3, dall-e 3 review, dall-e 3 first impression

Id: hZ-eEcnS_CA

Length: 14min 29sec (869 seconds)

Published: Fri Oct 13 2023