chat GPT has a brand new way for making images i've been using it all day i have 10 examples I want to show you 10 use cases this is called 40 image generation so it's a native way for chat GPT to make images it's no longer using Dolly so I've been using Dolly for a long time dolly 1 Dolly 2 Dolly 3 those are now removed from Chat GPT and I want to show you this new way for making images which is a massive massive improvement from the old way now this is available to all the paid versions like the plus the pro the teams version it is not yet available in the free account though okay let me start showing you some examples and some use cases the very first use case is making a product mockup right here so here's the prompt that I use and I'm not going to read through the prompt but I gave this prompt exactly what text I want exactly where I want the text the details of the box over here and all kinds of different ways I want the colors to look and look at the details of this thing here's exactly how I spell chocolate in the prompt that I gave it and it followed every single ounce of that prompt perfectly to a T the next example was creating a mockup for a website the entire banner so I wanted to create one for a resort that is trying to get people to book a vacation and I gave it all the information here for the menu that I want i told it I want center text that says this everything about the website okay look at the details of this thing and here's the really cool part let me go back i got this in square because I didn't tell it the size so I followed that up and I asked it to create this in 16 by9 format for me and it just kind of stretched it out every time you click on one of these right down here you could ask it for a revision or to remove something or to replace something and it will go ahead and do that so if I just wanted the background and all this removed I could do that if I wanted to change the text right here I could go ahead and give it a prompt right down here to do that okay for this next use case they actually created a very realistic looking photo with a lot of text okay so if you need something with a lot of text which couple of them I showed you with a little bit of text but this example they had they gave a ton of text here and described this whiteboard and this is the image they got out of that reflecting the Bay Bridge here in San Francisco and the text was exactly the way they had in the prompts but a lot of times I need to test these out on my own and here's the version that I got so the person's in a little bit of a different spot but it did not say that in the prompt but it got all the text right here and you could see the Bay Bridge here in the reflection the person taking the picture now let me show you how far this has come so we had this prompt book here for Dolly 3 and these are different prompts that we had that generated these images so I took this exact same prompt from Dolly3 which the image generation platform we had literally two days ago inside of chat GPT was Dolly3 that's now removed and replaced with this native chat GPT image generator and this is the prompt I took so here let me zoom in this is what this gave us this exact prompt with dolly 3 and this is what we got inside of chat GPT with 40 image generation here's the next one I had this is a white photo inside of a grand library but Dolly always made things look a little bit unrealistic right it didn't look like a photograph it looked more like an illustration right but look at this one now this is so much more photorealistic compared to what we were getting with Dolly 3 okay I tried it with this one from Dolly 3 prompt book this was a close-up photograph of a vibrant butterfly and more details about the flowers so this is Dolly 3 here and this is the new image generator inside of Chat GPT i mean this is clearly a world apart now Dolly was not at all good with making any type of portraits or any type of human faces so this one right here a close-up photograph of a young athlete i mean there's all kinds of issues it's much better illustration than any type of a realistic photo but I was doing this is a photography prompt book and this is the best I was getting out of this one okay so yeah this is what we got now this is what we had before it's I mean it's not really comparable i'll just show you one more example here's a vintage portrait of a person where I said fulllength portrait it didn't quite even follow the prompt that was over here in Dolly and here's what we got and it follows the prompt a lot closer it looks a lot more authentic as a vintage photo okay now that I've compared it with Dolly 3 let me show you some more practical use cases here i make a lot of YouTube thumbnails obviously so I wanted to turn this into a YouTube thumbnail i said make a YouTube thumbnail of this person cut them out of the background put a techy and blurry background instead and have him hold a glowing open AI logo okay so that's me that's the prompt that he followed almost exactly but I don't look quite like me i mean it's in the same world but that's clearly not me right so then I said "Well no make it exactly the same person." Well that's a different person so it was pretty close but you could tell it's not quite me yet so I think for thumbnail generation it's not quite there yet but I tried to turn myself into a wizard i said "Well make me a wizard with a wizard hat and a robe and a magical expression." I wanted to see how it kind of thought about that if I said magical now it's more cartoony looking and it's still not quite me but it's getting a little bit closer when it looks a little bit more not like a photograph on this one I said replace all the logos with the top seven AI company logos so now this is a little bit more than just image generation it has to go find logos figure out what the top seven are and it did a pretty good job it created eight logos but you could see anthropic open AAI Google this hugging face midjourney right it got few I don't know what a couple of these are actually but it kept everything else the same again I don't look exactly like myself it's in the ballpark of myself but not quite right but I said hey you cropped it sometimes it crops the left side it did not fix it in that case and it kind of made it a little bit worse it's pretty close though to actually be able to generate entire YouTube thumbnails the fact that it could cut someone in the background generate a new background find logos get the text right that is a huge leap now the next use case is coming up with infographics because it's so good at text i asked Chat GPT to create a prompt for me and I kind of described what I was looking for i was looking for an infographic to show evolution of video games and first chat GPT gave me this prompt where I put these video games including some stuff that's coming out later nextg cloud console which is a concept so some real ones some concept ones now look at the details of this thing the amount of text that he had to get right he got almost all these right by the way I think there was a little bit more to this timeline but to fit it here it looked like it started from the beginning and took us all the way to this concept next gen console but I mean you even got the shapes of these things like look at the Nintendo right here now the next prompt I tried was to create a meme so I asked ChatP to make one up for me he created this prompt for me and the first time around he cropped it so you couldn't quite get it and I said "Hey you cropped this." I just clicked here i said "Hey you cropped this give me the text without cropping it." And he actually fixed it for me it kind of stretched it out so the text fits perfectly right and it's ready to post literally I could click here download it download it to my computer and upload to social from there now the next use case is changing the style and look how well this did i gave it this thumbnail i said "Turn the person on the left to a cartoon but keep everything else exactly the same." And look at this it kept things exactly the same it kept the background even kept a little bit of that OpenAI logo here that I had but it turned me into a cartoon that's incredible now the next one is for creating graphic markups i actually wanted to see if you could mimic something famous like the cover of Time magazine so I said create a high-end Time magazine cover featuring a confident individual and in the prompt I said photos to be inserted and I forgot to insert it i'll show you something interesting happened looking directly with a visionary expression and then I said surround that person with the logos of the top AI companies this is what chat GPT just chose for me when I was crafting the prompt and then here is the exact text I wanted okay let me scroll down okay look how well it copied that prompt here it's an exact Time magazine cover it put the logos here surrounding the individual that I asked for it put the right text over here and it put a person right over here which I didn't actually intend this to be me i was going to put someone else here but it actually looks a lot like me so I don't know how it figured that out because I did not include any photos it just said that placeholder said something will be inserted it did not do that but that looks a lot like me so I don't know exactly what happened there and then I was like "All right well let me try let me give it this picture and say put this person but have them crossing their arm." Okay everything is the same looks the same it's surrounding me but it did not get my face right i guess it kind of looks like me but again different person it did follow the crossing the arm part of the prompt and it got the rest of it right too so and there were a couple limitations that I wanted to point out that I noticed so Dolly 3 actually got this one right back in the day create a wide photograph of a person standing in front of a bright sunlight and using that to kind of create that warm glow around them well every time I tried this for some reason the sun goes through the person so that's not the way that should look so there was a couple limitations I came across very rarely though did it do something like this where it wasn't following my prompt the other limitation was from time to time it would crop things and the fact that it just couldn't get my face exactly right which for my use case of creating YouTube thumbnails that limitation is big enough that it kind of takes that out of the equation i still will have to do it the old-fashioned way of using Photoshop but it's getting a lot closer now and if you're trying this out for yourself inside of your chat GPT accounts this is a lot slower than Dolly used to be it's a lot slower than any image generation platform i use ReCraft I use Midourney those are a lot faster so maybe that's just related to the fact that this came out about 24 hours ago and I've been pretty much using it all day and I'm inside of my Pro account is even a little bit slower inside of my team's account so you do have to be patient with it that's pretty much most of the day here that's about all I was able to generate so far but hopefully that does speed up now as soon as I get more time with this I'll update our prompt book and I'll make some updated videos about how you should prompt this new model here thanks so much for watching i'll see you on the next video