Transcription
Did you know that with the new version of chat GPT, you can recreate wild thumbnails like this, completely design a bedroom with an image like this and turn it into something like this? You can make animated parallaxing backgrounds like this and even turn yourself into your own action figure like this.
And even though Sam Altman is saying the team over at OpenAI isn't sleeping because everybody's using these image generators so much, they've now rolled it out to even the chat GPT free plan. So even free users of chat GPT can generate crazy stuff like what you just saw in this video. I'm actually going to break down how actually insane this new release from chat GPT is because I don't think most people realize how truly disruptive it actually is to so many people and so many industries. There are so many things you can do with it. I've rounded up something like 50 different ways that you can use the new chat GPT image generator, and I think some of these concepts are going to blow your mind. This is going to be a fun rapid-fire session where I show you some insane uses of this chat GPT tool. I'm going to start a little bit basic, but it's going to ramp up as we go. This is going to be a fun one, so let's get into it.
So first off, in order to use it, it's really easy. Head on over to chatgp.com. When you're logged in, make sure you're using chat GPT 4.0, and then just tell it what you want. Just like any other image generator, you can generate pretty much anything you can imagine. For example, "create an image of a wolf howling at the moon in this style of GTA 5, with the text 'subscribed to Matt Wolf' on the image." Now it's not the fastest image generator in the world, but after about a minute or so, I got this image back, and the text is perfect, and the style is exactly like I wanted it. Here's a thread from an unnamed Akash here with some other really impressive images. Here's an image of Miaki that's half Ghibli. If you want to know the prompt, you can go ahead and pause the video; it is a longer prompt. "A grungy analog photo of Lady Gaga from a luxury lifestyle flash photography, unedited." Here's one of the Pope stealing some wine. Go ahead and pause the video to see that prompt. "A woman holding up some chipped fingernails to the camera." Remember, like last year, AI couldn't even generate hands. "Portrait of a realistic Queen emerging from a playing card." Again, it's also really good at putting text in the images. Here's one of somebody writing a ton of stuff on a whiteboard; you can even see the reflection of, I think, the Golden Gate Bridge in the background, and all of it looks really good.
Creating images from scratch is super impressive with the new version of chat GPT, but it's only scratching the surface of what it's capable of. Now you're probably aware that it can already restyle images, where you give it one image and then change it to a completely different style. You've probably seen the internet get flooded with Ghibli-style images; that is another thing that this new model of chat GPT is really good at. However, instead of showing you a bunch of Ghibli-style images, here's some other ones I've come across from people. Red Devil and Y here says that "photos in the style of cyberpunk hit hard." This is one that they restyled in the style of cyberpunk. Holly on the farm says, "I've been feeding photos from my travels and messing with different painting styles," so I'm guessing this is like the original photo here, and here it is restyled in like different art forms. Here's a few more from that same thread. Here's one that I found on Reddit where they took this sort of anime screenshot here and said, "Can you please turn this into an epic pixel art?" and it restyled this anime image into a pixel art image. Here's another one that's in, I guess, this animation style here; I don't know what you'd call that style, but they said, "Please generate it in a modern anime style," and it restyled it into a more modern anime. Ralph Lentes here converted himself into a Funko Pop style. I liked converting myself into a South Park style, a Minecraft style, a pixel art style, and a boxel style. I've also been a big fan of this GTA 5 style where I took this photo of Padres players and converted it into a GTA 5 style image.
It's also really good at taking an initial image that you give it and then editing that image. So here's the hero Shep here; they gave it this image with this character over on the right side of the screen and gave it the prompt, "Can you put the person in the middle of the field in the distance, halfway to the castle?" And then it basically generated the same image but put them out in the middle of the field. They then fed this into an AI animation tool, Cling 1.6, with a starting and end frame and animated it. Here's some of the edits I've done when messing with thumbnails, which I'll even talk about a little bit more in this video a little later on. But I had this image with me holding up these little AI pin things, and I wanted to add the text "worth it." The first time it didn't do perfect; I tried it again, and the second time it kind of nailed it. I had this image that my thumbnail designer made for me; I wanted to change the text to say "uh WTF." It easily swapped out the text for me. I then wanted to convert the same image to me with a frowny face and change the text, and it made this version; looks pretty good.
It's also really good at combining images. Gizm Act Dag here created this image on the right by combining this image with this image here and ended up with this composite of the two. They gave it the prompt, "Insert this plate into the center of the right image; match lighting and perspective." Here's an example from Reddit where this person took this anime character and this scene from like a co-working space or something and said, "Can you edit my image so that the anime girl is sitting in one of those chairs?" And it combined the anime girl with the image in the background and made her sitting on the chair. And then I was screwing around by taking two images, one with me on one side of the camera, one with me on the other side of the camera, and then asked it to combine the images with an explosion in the background, and I did it multiple times, and then I Ghibli-fied it. You could turn drawings into photos. Sand Cop here gave it this drawing and then said, "Make her real; Polaroid photograph with film grain effect," and got this Polaroid-looking image out of it. Here's one where they took the famous meme of the person crying and the person just sort of staring and made it realistic-looking. Here's another one where somebody took a meme and converted it to a real-life-looking image, and there are tons of examples of this style of image getting converted into a much more realistic-looking image.
You can even take handwritten notes, like Mustak Bal did here, where he just put text on a piece of paper: "Banner showing happy graduation, a woman standing wearing a red suit, a man standing wearing a blue suit," and it converted it into this image just based on a picture of a piece of paper with some text written on it. Did you know that chat GPT can also remove the background on images and turn them into a transparent PNG? So this image I made earlier, if I just say "remove the background so it's just the wolf and the text, make it a transparent PNG," after about a minute it did exactly just that. I could download this, and it won't have a background, so I can overlay it on anything. You can also ask it to remove things from an image that you don't want in the image. Here's a Ghibli image that I made; "remove the robot on the right." I get pretty much the exact same image back but without the robot on the right. It works with real photos too. So let's say I have this image, and I want to remove this light on the right side. I could take this image, throw it into chat GPT, "remove the light on the right side," and we get pretty much the same image back but with the light removed on the right side. You could even change the background and put people in a different environment. "Change the background; put her in a rainforest," and now we've got the same woman holding the camera but in a rainforest.
It can also colorize black and white images. Here's an image from Broski here where he has this black and white image, and he told it to color the image, and we got a color version of the image. I could even take an old black and white image like this, toss it into chat GPT, and say, "Colorize this image with bright vibrant colors," and I get a vibrant colorful railroad crossing sign. You could create professional headshots, like Lonus Ekenstam. He gave it this original photo, and it gave him this headshot back, and there are some other examples in this thread of other people who have generated some headshots. It can make really good e-commerce product shots. So here's a Reddit thread where somebody gave it this cap and said, "Can you put this cap on a man at the beach; realistic please?" And it took this O'Neal cap and put it on a man. "Can you make this surf skate be resting against a wooden wall in a room that looks like a cabin, etc., etc." Just like that, it got an image of that same skateboard against a wall. Here's an example of somebody taking this cat meme image and putting it on a T-shirt, and here's another example of it on a different T-shirt. So if you're selling T-shirts, you could instantly create mockups of your image on a T-shirt. Here's an image of a razor from Ali Kesi, and here's a product shot that could be used on an e-commerce store. Here's another example from Salma here where she was trying to create this grid product shot: "white background, product in use, product benefits, product testimonial, product texture shot." She gave it this input and got this output from it. Here was her exact prompt; you can see she gave the little grid here along with the nail polish and said, "Using the grid provided, create images for each style for the nail polish product." This was the output.
Now this is a pretty niche use case. My dad runs a shutter company, and he asked me if it can make mockups of what the shutters would look like. So I tried it: "Created an image of a plantation shutter with 3 1/2 in louvers, two panels, 60x60," and it created a really close-looking mockup. Now the type of shutters that my dad makes have like a tilt rod thing down the middle so you can open and close them, so I said, "Add a tilt rod in the middle of the louvers of each panel." Niche terminology, niche product, and it actually nailed it. It can be used for virtual try-ons. Here's a thread from Halam Al-Rishi. He took this photo of a woman and said, "Let her wear this jacket; same pose," and uploaded an image of the jacket and got an image back of that same woman wearing the jacket. Here's another example of a woman in this pose here; it said, "Let the model wear the jacket; same pose," and got this woman wearing the jacket. You can actually go the opposite direction and take images of people wearing clothes and get it to extract the clothes from the image. Somebody took that image here and extracted this T-shirt from it. Here's another one where somebody took this image from The Hangover and extracted the wardrobe from it.
It can be used for trying out new hairstyles. Here's what Masha had to say here: "I used it to try different hairstyles, then print out an instruction infographic and mockup for my hair stylist." Look, it even came with instructions on how to cut and style the hair in that way. I believe you can also do this with real images as well. If I go to chat GPT here, upload an image of myself and say, "Give me a mullet," let's see what it does. Here's roughly what I'd look like with a mullet; maybe I should rock it. Some of the most impressive stuff I've been seeing has actually been people redesigning the interior and exteriors of their homes. Paula and Mike here posted this: "The first picture I had chat GPT remove everything in the room so I can see how it looks when I move in, and then I shared the kind of bed I would like and had it added to the room." Here was the original image with a bunch of stuff in it. Here's the second image with the room completely emptied out, and then they started adding their own style in to see some options of what it could look like with their own design tastes. Here's one from Justin Strikes: "My parents-in-law were curious about how their house would appear with the garage door, side walls, and dormer painted dark gray." I'm guessing this is the original image here, and here's a version with the house painted. Here's another original image it looks like, and here's an altered version with the paint done on it. Here's one that I found on Reddit that was really interesting; they took a picture of their gazebo and their backyard, and then they sketched in what they were looking for, and this was the result here. Here's another angle; they sketched in where like a counter would go, and this is what it generated for them. Here Shingala gave it this image: "Turn this into an aesthetic house with good lighting and professional photography." This was the redone version of it. EP here posted this; it looks like they probably found it on Reddit. Here was the original image of a bedroom, and then they redesigned it to clean it up and make it look a lot better. Here's another example from George Martin Jay-Z here where he took this view of the balcony and sketched out where they wanted things to go, gave it some images of chairs and a table that they wanted to see, gave it an image of a plant, and ended up with this mockup showing what their outside balcony could look like. Here's another final example; they created a little mood board here with this wallpaper, this green color, a bed, a lamp, a nightstand, etc., and it mocked up this look for a bedroom for them. This one came from Jenck Man here.
And speaking of mood boards, Rory FL here shows off that it's really good at making mood boards. Go ahead and pause and read this prompt if you want, but they actually put hex color codes in here for the background and the puffer jacket and everything, and it created this mood board with all sorts of images with those colors from the palette that they suggested. I actually asked chat GPT, "What's another interesting way that's sort of outside the box to use this tool?" And it said, "Create educational infographics to learn about historical events." It actually gave me this prompt to try: "Show Newton conducting his prism experiment as an infographic." Here's what it gave us back: "Newton's prism experiment; sunlight going into the prism; the rainbow coming out; Newton holding the prism." chat GPT can also help you illustrate children's books. Here's one that I found on Reddit here from Robot Shakespeare about a book called The Very Hungry Alien. All of the illustrations in this book are in a very similar style, so it's really good at that consistent styling. We can see here's the first image with the actual text for the book; here's the second image, and it all has that children's book styling to it, and it all feels like a very cohesive story about an alien eating people. But if you want to write a children's book, chat GPT can definitely help with that.
Maybe comic books are more your style. There's been a lot of examples of people creating comic book pages using the new chat GPT. Here's one from Daria where you can see here's some of the comic book pages that they created, and again, very consistent style across all the pages. Here's another comic book page from Kyle Balmer here; completely different style. Here's one from Red Devil; this looks like a GTA-style comic book; another completely different style on the comic book theme, but still looks like a comic book nonetheless. I mentioned infographics earlier; here's a Venn diagram that I created of how I sort of see myself: entrepreneur, creator, AI tech nerd, with me in the middle. Here's one that I created with how a neural network works. If you look closely, there is some typos, and it's not perfect, but I also gave it a very, very basic prompt: "Create a colorful, beautifully designed infographic that explains how neural networks work." I could have steered it a little bit better with a much more in-depth prompt. Here's one about how a diffusion model works, and here's one that I made about how reinforcement learning works. Super, super basic prompt here with a little bit more depth; I could have dialed it in even better. Luckily, other people have made much better infographics than I have. Ali Miller here made this infographic about how to live in New York and then actually took that infographic that she created and had somebody hold up a sign with that same infographic on it. Here's another infographic about how to take your cat on a leash walk, and then she had it injected into an image of a cat looking at the instructions. Here's one from here showing how to make a pho recipe with all of the ingredients, what you need, and then what the final dish looks like. They even shared the prompt here, so if you want to go ahead and pause, you can read that whole prompt, and you can create something similar. Here's another one from Eay; I'm not sure how to pronounce a lot of these names, so I apologize if I'm butchering them, but it's another recipe infographic with all of the ingredients with the final dish at the bottom. If I hover over the alt here, you can actually see the prompt that they used. And finally, here's one that TBO shared where they uploaded a text document with all of the information they wanted in the infographic, and it spit out a pretty decently designed infographic with those details in it.
Let's move on to website and design skills that you can use chat GPT for. You can create entire website designs. This is a design from Jay here, and you can see the entire prompt that he gave, and again, this was the site that it generated. Here's one from the Samurai emphasis on the AI where they said, "Create me an image of the first section of a website UI of a highly modern website of a SaaS app for a YouTube thumbnail maker." You can see the design that it actually spit out for them, and once again, it's a solid design. Here's another one from Yoko; they created various landing pages for Andre and Horowit here. Here's a more anime-fied version; here's another version, and here's another version. And what's really cool is you can actually take these screenshots, and if you head on over to DALL-E 2, you can actually upload the screenshot and say, "Make this a functional website," and just submit it with nothing else but the screenshot, and it will actually go off and code that website for you, and you get a functional website. Now the original image that I submitted actually had text in the image, but if you were to go to chat GPT and say "remove the text," you wouldn't have that duplication. So you can see it's actually got the real text, but then there's also text in the image behind it, but this website is coded and functional, and you can actually download a zip file with the code for this website. And here's a bonus website; this person, Riley Goodside here, actually made a fake screenshot generated by chat GPT of a Wikipedia article about the screenshot itself with a copy of the screenshot in the article. They even shared the alt text prompt; it's a pretty in-depth prompt here.
The new chat GPT seems to be really good at creating ad images. You give it a product image and describe what you want that product image to look like, and it will spit out an ad for you. Here's an image about a selfie booth that was generated. Here's one from Sola about sparkling water: "Venom Sparkling Water." Here's another one from Selma about a hair care product: "Tame Your Mane; before/after." This was created with GPT 4.0, and here's a screenshot of the actual prompt that was used. Here's one from here: "Create a Mad Men-style print ad using this extremely popular Indian product." This was the ad that it created. Here's some more examples from Ali: We've got this David protein bar: "The ultimate toolkit for male confidence." "Hunter and Gather Collagen Peptides for great hair." "Anti-dandruff shampoo; introducing the power couple for your scalp; battle the flakes." It can actually create icon packs for you that are all in a very similar style. So if you're building a website or an app, you can actually have all of the icons look fairly similar. Here's one example from Thomas Bre; here's another example from Kuba of another like little icon pack that they created. Here's another pack that was made by Jack here, and here's another icon pack from Sashin here where they gave this original image and then had it create more images in a similar style, and there's all sorts of different icon examples here. Apparently, this version of chat GPT can also create repeating backgrounds as found by M from scenario.com. They made this brick repeating background here. Here's some more bricks in different colors. Here's the prompt structure that they suggest, and here's another example. Here's a repeating pattern, and you can see that this is what it looks like when they sort of tile it and get it to repeat. Here's another one of like cracked sand; there's what it looks like when it's repeating. You can even get chat GPT to create fonts: "Create a font based on the style of this question mark; you can ignore the messy white parts of the image." This was just an artifact of cropping, and then it made a font for them with these. Here's another example from Sero: You can extract fonts from any graphics using chat GPT and build your own typeface. Here's the original graphic; here's the font that it created from it. Here's the "Severance" text, and here's the font that it came out of it. Here's another one, and the example font that it came out of. Although looking at it, I mean, these Rs aren't quite the same, and neither are these Es, but I mean, I guess it's close-ish. Here's another one from this "Girls Don't Cry" text, and here's the font that it generated. Here's some more fonts that the Beused Studio shared over on Threads that they managed to create with the new chat GPT. Here's even more.
It can create logos. Here's one from Caesar Flores: "I asked chat GPT to create a logo by merging a dog and the letter R," and it made these logos for him. Here's another one; he also made this logo with chat GPT; another dog image. You can create business cards with it. Sola here said, "I just made this business card for myself; super crazy." Here's one side of the business card; here's the other side with the text and their contact details. Now let's move on to some of the even crazier stuff I've seen this do. It has been making some wild thumbnails. Here's one from Colin and Samir where they took this thumbnail from a MrBeast video and said, "Replace the subject in the thumbnail with me," and here's a picture of Samir, and this was the image that it gave back. He then said, "Change my shirt to white and add rips to it," and it made those changes, and we can see here's the original MrBeast thumbnail, and here's the Samir version of the thumbnail. Here's another example that Lane shared here where they drew this sketch with a dude hitting a baseball bat, and they added text on it: "Jersey, baseball bat, light, trees, 1000 mph, blue sky, flames, fire, etc. Turn the sketch into a hyperrealistic YouTube thumbnail, 1920x1080. The only text that should stay during the creation process is '1000 mph' and the arrow right next to it; the rest of the text are instructions about details for you to follow," and then it made this thumbnail. Here's another example that EP shared; they gave it this image up here; they gave it this prompt along with it, and they got this output. Here's another example from EP: "An astronaut with a cracked Soviet Union flag, hand on the ground, scratched glass, the moon, left to die." This is the output. Another weird issue that chat GPT still does have is every once in a while it does crop off text, so sometimes you need to get it to like sort of readjust to get the text to work right, but I mean, for the most part, it did a really amazing job. And here's another one that EP shared; there was a photo of this guy here: "$400 million CEO, office background, money on fire, hand with a bill," and it converted it into this image.
I already showed you some of the stuff I was trying to do with it, and it's been really, really helpful for me. And again, you can edit thumbnails with it as well. So if you have a thumbnail like this and you want to change the text on it, it's really easy to do. You want to change to a frowny face; that's easy to do as well. It's also really good at creating movie poster styles. Sid here created this poster with these two shooters from the Olympics with the text "John Wick" on it. Zaye KH here actually took his own image and the Terminator poster along with this prompt and created a Terminator poster with his own face on it that says "The Content Creator." And here's one from Joanna where they used chat GPT to make a poster for their Death Valley movie; I mean, it looks like a poster you'd see at a movie theater. chat GPT is also really good at generating storyboards for you, so you can give it a detailed sort of script of what you're trying to create and let it actually generate the storyboard to go along with that script for you. I even recently used chat GPT to add some B-roll into one of my videos. If you've watched a lot of my videos, you might have even noticed this. I took this image of myself; I needed something on the screen during a time where I said, "You probably know the pain of juggling multiple tools." I didn't know what to put on the screen, so I had chat GPT generate this image of me juggling with that text on the image. You'll actually see that pop into one of my videos. I then had another shot where I said, "You probably bounce between five different apps." Once again, did the same thing, gave it an image of myself with a prompt, generated this image, and used that as part of the B-roll in my video. You've probably also seen a lot of videos like this recently where it's like the Ghibli style overlaid on top of a video. This is actually a podcast from a16z. If I unmute it for a second, you'll see it's actually synced up to her audio: "Utility, most delightful, most helpful to a consumer user," and they created this whole podcast with this overlay on it. Here's another one that kind of went viral on X here with only, you know, 24.8 million views where they took the cast of Severance and did it in various different styles: "Rudolph the Red-Nosed Reindeer style, Don Bluth style, Fleischer Studios 1930s rubber hose style, Pixar style, toy animation, I don't know how to pronounce that, Cow and Chicken Cartoon Network style, retro futurist style, and this Syvan CH style," and basically what they did was they just took a video, took the first frame of the video, and then had it restyled and then used Runway to make the rest of the video in that style.
I know this video is getting long already, but I want to demonstrate this, so if you want to do this yourself, you can. So check this out. Here's a video that I recently did; it's an interview that I did with somebody at Nvidia, and let's go ahead and restyle this in Studio Ghibli style. Why not? I will take the first frame of this video here; I'll just copy the image, jump into chat GPT, paste the image of this first frame here, and then say, "Make this in the style of Studio Ghibli." Now that I have my Ghibli-fied image, I can go ahead and download this, and if I head over to...
Runway ML: We'll go to Act One here, and then I'm just going to go ahead and upload like 10 seconds from that interview that I did. Toss that right there. And then, for our character reference, I'm going to go ahead and upload our GIF-ified image here, and we'll go ahead and generate it. And then we get a Ghibli animated version of the interview that I did. It kind of froze me here, but that's how that's being done.
New AI world over on X shared that they were actually able to get ChatGPT to generate every frame of an animation. So it generated—it looks like 12 frames here—and then we can actually see the output animation that it generated. The new ChatGPT can also help with all sorts of areas of game development. For example, George Martin Jay-Z here shared that he created these like character sprite sheets with ChatGPT, as well as these various game assets. He then actually implemented these graphics into a little video game that he created here. George also shared these character sheets that he created and some other game assets and items from within the game.
I wanted to test it myself, so I created this little wolf sprite sheet here. I gave it the prompt: "Create a character sprite sheet for a 16-bit video game about a wolf. Show the wolf from different angles, as well as some collectible items for the game." And it made this little sheet here, and even made the background transparent for me. You can make a parallaxing background in just like this. Here's the prompt that he used. You can go ahead and pause it if you want to see it; it's a pretty long one, but it generated the three sort of layers of this parallaxing background. He then used Gemini 2.5 to create a little script to make the images parallax.
You can turn yourself into a game character. I actually uploaded this image of myself and gave it the prompt: "Turn me into a 3D video game character in a T-pose." It made this 3D character of me, and then I was able to take it over to Trip3D and actually turn it into an actual 3D object here. And not only that, but I was able to rig it up and animate it so that I can make myself walking or doing a somersault, and I can look at it from different angles as well. I can even make myself fall down.
I want to try this with the voxelized version of my little character that I created here, so I took this one over to Trip3D and turned this into a 3D object as well. This stuff is super fun. Now I tried to actually animate this one, but the rigging didn't quite work in Trip3D. I just kept getting this unknown error on the server side. I think maybe it just doesn't look humanoid enough; I don't know. But I was able to pull this character over into Luma Labs' Dream Machine and create a little animation of this character walking. It's not the actual 3D object walking, but it looks pretty good. I was able to make the animation of the character jumping as well, and apparently when I jump, I do the splits a little bit.
You can turn pixel art into 3D art, like Ox SJ did here, where he took this little pixel art monkey and turned it into a 3D version of the monkey. Here's a post I found on Reddit where somebody took this like Pop-Tart cat meme and turned it into an animate style of the Pop-Tart cat meme. You can take 3D sketches and turn them into 3D objects, like Daria did here, and she actually took this and threw it in an app called The R App, which I wasn't previously familiar with, but this turns your images into like little user interfaces. So she started with a line drawing, turned it into a 3D object, and then converted that whole thing into this UI that you see here.
CLA Silver shared on X that you can actually use ChatGPT to change the reference angle on an image to create PBR textures and to put it into AI-to-3D sites. We can see here that she gave it this texture: "Please make a transparent PNG of a realistic teapot," and it made this teapot with a similar design. She then said, "Please show a top-down view of the teapot," and it made this version. "Create an image normal map for this if it were a sheer and velvet embroidery curtain," so it took this and turned it into this map here.
You can reimagine games in different styles, like Broski did here. Here's one that I think might be like Red Dead Redemption, but reimagined in more of like an old-school style. Here's another one that I also think is supposed to be Red Dead Redemption. They also made like a pixel art version of Elden Ring, which I actually kind of wish I could play. I love Elden Ring, by the way. And here's another rendition of Elden Ring in like a pixel art style.
I wanted to try this myself, and I used to love the game Monkey Island, so I said, "Create an image that reimagines the original Monkey Island game, but in the style of GTA 5," and it made this image. And then I said, "Create it in the style of Breath of the Wild," and it made this image. And then I asked it to make it in the style of Dave the Diver, which is like a modern-day pixel art game, and it made it in this style. Super fun stuff.
You can use it to create board games, like Ru and Sasha's Happy Days did here. They created Candyland with a dark twist with Marilyn Manson in it, and they created a Willy Wonka version of Chutes and Ladders. But what this is mostly getting used for these days is sort of fun meme-type stuff, like the type of Broski is doing here, where he turned an image of a cat into stickers of a cat, and like Yoko did here, where she made some Andry and Horor Wit stickers, as well as some a16z stickers using the new ChatGPT. It's a great meme generator. Somebody took this photo of Sam Altman and this famous image here, and they put Sam Altman's face on the image. Ethan Mik used it to create this meme here.
LLM image generation: Now, thanks to Google and OpenAI… Wait, you're telling me it didn't already work like that? Yeah, it's kind of getting to the point now where this image generation is working the way it feels like it should. You just type a prompt and actually get back what you thought you should get back. Flare here shared a "Dude Wears My Car Tamagotchi" version. You can also create fake magazine covers, kind of like Here did, where she put herself on the cover of Life magazine, and Dirk did here, where he put himself on the cover of Mad Magazine. I wanted to test it, so I put myself on the cover of Time. Why not?
Now, this is something I've been seeing everywhere. This is like the new Ghibli-fication; it's the action figure occasion. Here's Julia McCoy as an action figure, Michael Jackson from Thriller as an action figure. Marcio made the Predator. Here's an Elon Musk action figure, a SpaceX Elon action figure. We've got action figures for every character from the show Severance here that all look pretty dang good, actually. Maybe the faces are a little bit too realistic, but I don't know, the show's already kind of creepy. Christopher here even took this and had it animated into somebody looking at the packaging using Labs. Here's a Kendrick Lamar action figure, a Gnarly XYZ action figure, a Snoop Dogg action figure. These things are everywhere right now.
You can turn yourself into a Greek statue. Here's the prompt: "Turn me into an old Greek statue; background is full white." That's the whole thing. Then you can turn it into a 3D object, like I did with my game character using Trip3D or Meshy or this Hugging Face Space that they shared here. Here's some other random ones from Flare where they made Beanie Babies out of the movie The Craft and Beanie Babies out of… that's TLC, maybe. Here's some Happy Meals that Flare made with celebrities on the front of them.
I've been seeing these more and more as well: these like unhinged posts, like the one Min Choy shared that just has all of this crazy, unhinged writing all over it. Here's another one from FFR here with all of this crazy, unhinged writing all over this tweet, and there's all sorts of other examples inside of this thread here. You can also give it nonexistent, random words, like Plany the Liberator did. He asked it to generate a portrait of a CL, then that's what they got. Here's a Growen Bops and a Val Piters and a Mler. I wanted to try this, so I had it create an image of a Flagger Snort and a Gloopy Lar.
Now, so far everything I've talked about has been sort of like fun and memes and stuff. For your business and advertising and YouTube thumbnails and, you know, creative graphic design type stuff, but this could also be used for more, you know, nefarious reasons. For example, it turns out this is really good at making fake receipts here. So when it comes tax time and the tax man wants to see some receipts, well, are they going to be able to actually trust the receipts anymore? In fact, TechCrunch put together a whole article here called "ChatGPT's New Image Generator Is Really Good at Faking Receipts," with a whole bunch of examples of other fake receipts that people have generated using this ChatGPT image generator. It apparently can create fake autographs here. Here's one from Andrew of Michael Jordan that's been signed, meaning some of this memorabilia there might be even more fakes flooding the market. And here's one from EP of fake damage where they actually took an image—image—of a product inside of an Amazon box and then had ChatGPT make it look like the product was broken so that they can get a refund on the product.
Now, I'm not showing you these things because I think these are things you should go and do. I'm showing you these things because I think you should be aware that this kind of stuff exists now and that you could no longer believe your eyes. You need to sort of question every image you see, and that's pretty much what I got for you. There are so many different ways that you can use this tool; it is just mind-boggling. I actually posted over on X that I rounded up 35 different ways, and this video was going to be those 35 ways. However, you guys shared tons of additional ways that I hadn't thought about yet in this thread, and now this video ended up being well over 50 ideas. So thank you so much to everybody who commented on this thread and helped contribute to this video that you're watching.
And before I wrap up, here's some quick tips. Here's one from Lonus where he claims that Sora equals less nerfing. He was having some trouble generating images of his family directly inside of ChatGPT, but apparently if you go to the Sora user interface, it refuses to make images less often, and he was actually able to get it to generate a lot more images with his family in them. Also, if you go over to sora.com, you can actually see a feed of all sorts of images that people have generated with this new model. So if you need some more inspiration or ideas that we haven't thought of yet, just scroll the feed, because you'll get tons of ideas. And if you do have ideas that we didn't talk about in this video, make sure you leave them in the comments below so that other people can learn from what you have to offer as well, because I know there's ideas that I didn't even get to in this video, despite how long this thing was.
Here's kind of my final thoughts on all of this and where this is all headed. I actually feel for designers and artists and people who create things like this for a living. I know this is probably a really tough era to be going into, and I even worry about it myself—that AI is eventually going to be able to create content way better than what I can create. I mean, if it can't already. However, I do believe that a lot of these things take like a good eye; they take skill; they take practice. Yes, the actual doing of the thing, the creating of the image and the creating of the art is getting more and more democratized, so anybody can do it. But the vision behind what you're creating, the ideas, the creativity—that's still required, and it still gets better with experience. I wouldn't know how to design a really good thumbnail like the baseball one we saw earlier or the astronaut. Those actually took a designer who—the vision of what the thumbnail could look like and mapped it all out with all sorts of text and annotations on the image to get that image. They likely had a background in graphic design and understood composition and color theory and things like that to get the image to what they wanted. That still requires skill and talent, so I don't think creativity is dead, even though you'll see that narrative a lot on social media. I think the ability to take what's in your brain and get it out into the world has sort of been democratized, but it's—it still requires creativity; it still requires vision. Even though I know I can do a lot of this stuff now, I still have a thumbnail designer that helps me because of his vision, because of his creativity, because of his ideas. And I don't mind if he wants to use AI to make his job easier; it just means I get thumbnails back even faster, and he gets paid the same amount for the work that he did for me, but he gets to do it faster, and I get excited about the end result because of his vision, not because of the ability to just press the button and produce it. I want to see a vision come to life. I've tried asking AI for thumbnail ideas; it's not very good at coming up with the ideas. I personally appreciate creativity and the skills that humans have, and I think other humans will continue to appreciate that and value that for a long time, even if a computer can now do it for you. But who knows, I could be wrong. That's just where I stand right now.
So thank you so much for tuning in. I really, really appreciate you spending the time with me today and nerding out over this stuff. I'm having so much fun playing with this. If you're into this kind of thing and you want to learn more about what AI can do and stay up to date on all the latest AI news, make sure you like this video and subscribe to this channel because I am constantly sharing tutorials and keeping you looped in on all the latest AI news. Also, if you haven't already, make sure you check out futurtools. This is where I share all of the cool tools I come across. I keep the AI news page up to date every day, and I've got a free newsletter where I'll just email you twice a week with just the most important AI news that you need to know and just the coolest tools that I come across. And if you sign up to the newsletter, you'll get free access to the AI Income Database, a database of really cool ways to make side money from using these various AI tools. It's all totally free over at futurtools. Thank you once again for hanging out and tuning into this video. I really, really appreciate you. This has been a total nerd-out fest for me, and I have had a blast playing with this and seeing all of your examples. Hopefully you enjoyed this video, and hopefully I will see you in the next one. Thanks again. Bye-bye.