Introduction to DALL-E and generative art - DALL-E Tutorial

- Given you're watching this video, you've heard of Dall-E. But many of you don't realize its name serves as the ode to Salvador Dali and Pixar's Wall-E, the first being a pioneer of surrealist art and the latter being a clunky robot who sought human-like companionship. It's an interesting comparison when you think about it, how AI is fitting in with humanity today. All right, so what exactly is Dall-E and this landscape of AI generative art? Well, let's start with some basics. Let's say you wanted to create a poster of an astronaut on the moon. If you type that request into a search engine like Google, it will search through millions, if not billions of existing images, finding all those which are tagged with raccoons or astronauts and the moon, and it'll present to you what it thinks are the best images it can find which contain as many of those keywords as possible. Dall-E is fundamentally different. When you type that request into Dall-E, it's as if you're asking an artist to create that image for you from scratch. Because like a human who grows up seeing all sorts of images of raccoons and astronauts and the moon, it will create a custom and original visual for you based on all these images it knows or has been taught. Think about that. Dall-E's like a human which has been given a massive archive of memories and it'll happily create for your original art by tapping into those memories to present to you something new. Now, there's several ethical concerns about this and we're going to discuss that in a full video later, so hold that thought. But for now, let me show you some of the examples of this incredible range of Dall-E. All right, so let's say we wanted to make an image of a lonely robot holding a red balloon. We call these requests or descriptions a prompt. And when you hit enter or generate, in a few seconds, Dall-E will present you four images. Now, these are looking pretty cool, but notice the style. We didn't really specify anything and it looks somewhat like a computer painting. But let's modify that prompt and now specify a pen and ink drawing. And as you can see, Dall-E will actually deliver on that. They look like they were created by pen and ink. Well, let's try another medium, say watercolor. And these definitely look like a watercolor painting. Or can it do maybe a realistic photograph? Well, it certainly can. I mean, it almost looks like we actually took a photograph of this little toy robot on the ground. We could even go with a little closer and experiment with macro photography. Well, what if I wanted to change the time of day and explore golden hour? Well, again, Dall-E has delivered and now I have four brand new original images. Now, in each of these, the robot kind of had a similar quality of like this almost vintage metal toy, but let's try something a little bit more modern and go with a black and white minimalist robot. Now, as you can see, I mean, this is not a Google image search. It's a unique visual creation based upon millions and millions of images recreated into something brand new and original. Now, one of the things I'm often asked is why don't my results look as good or what's wrong with my prompts? The simple answer is there's an art to prompting. First, you have to remember you're working with a computer. Yes, Open AI has done a phenomenal job in breaking down natural language, but you'll find the best results follow a logical description pattern. Here's the way I like to work. You start with the main subject matter, the thing that any viewer will first notice in the image. Then if there's anything else, like secondary elements, then add those. Then I get into the details, where the camera's located, the time of day, or the lighting, and the mood. And then lastly, I'll describe the style, such as pixel art or oil painting. Alright, let's give this a try. I'm wanting to create a happy felt cat sitting in a colorful garden. I'm going to add then the details, with a time of day, just daytime, and then the style. Let's go with just macro photography. We've seen that, so let's continue. And there you go. So you can see Dall-E delivered exactly what I asked for, and if you follow this approach, I think you're going to find you're going to get far more consistent results. So that's a high level overview of Dall-E and how AI generative art works. In the next video, we'll dive deeper into prompt creation, so you can hone your individual aesthetic style.