Generating images from text prompts - Firefly Tutorial

- [Instructor] There is something so amazing about being able to create images by just typing a few descriptive words, but that's literally all it takes with Firefly. In this movie, we'll look at how to use text prompts in Firefly to generate images. We're starting on the main firefly webpage, firefly.adobe.com. Here, we'll go to the first card the one that begins the text to image process and click to go to another page where there's a text box into which you could just enter a text prompt from scratch. But if you're just getting started with Firefly, you may like to get your feet wet by using one of the example images on this page. These are all images that were generated with Firefly. And if you hover over any one of them, you can see the text prompt that was used to generate it. And scrolling through here you can see the wide range of images that Firefly can create for you. There are examples of art in many different styles, photorealistic images, playful images, fantasy, backgrounds, composites, double exposures. Pretty much anything you can describe to Firefly can be generated. I'm really into dogs and so here's an image that caught my eye. I'm going to use the text prompt for this example image as a starting place for my own prompt. So I'll click on this card and that opens another page where you'll do the bulk of your work generating images in Firefly. At the top of this page, you can see four image variations that Firefly created from the text prompt that was on that example image and you can see that prompt down here in the prompt field. Now let's say that none of these four image variations works for you. You can have Firefly generate four more by just clicking this Refresh button. It sometimes reads generate or pressing Enter or Return. And if you want to see any of these images in a larger view, just click on it and you can cycle through all of these variations in the larger view by clicking the arrows to the left and to the right of the larger image. If you see an image you like but it's not exactly what you had in mind, for example, this one looks a little blurry, then hover over the image and go up to the Show similar button and click that. And Firefly will generate three more images in addition to the one on which you clicked. When you see an image that you like, it's a good idea to download it so you have it. You can do that by hovering over one of the smaller versions of the image or opening the larger version and clicking on the Download button in either place. Next, you'll see this prompt that's telling you that Firefly is going to add a tag to this image that will let people know that it was generated with artificial intelligence. And this is part of Adobe's commitment to promoting transparency around content generated with Firefly. So I'll click Continue there, and that caused Firefly to download a JPEG of that image. And here's that image in the Downloads folder on my computer. You can see that the text prompt for this image is part of the image file name, and that can come in really handy later if you want to generate more similar images and you just don't remember what prompt you used. So far, we haven't made any changes to the original text prompt, but of course we can do that modifying the text prompt to hone in more closely on an image that meets our vision. You can add words to the text prompt or subtract words or you can even select it and delete it altogether. Let's say that I want to get more specific about my Yorkie. I want 'em to be wearing a red sweater. So I'm just going to add wearing a red sweater to the end of this prompt and I'll click Generate, or I could just press Enter or Return on my keyboard. So the point is that adding colors or adjectives or more descriptive words can help you hone in on images that meet your vision. Some of the things you might add are actions or locations. So let's say for example, we want to see Yorkie dogs wearing a red sweater riding a bike on a mountain trail. We'll generate that and we'll get four more variations. And I think some of these are going to work. So I would go ahead and download the versions that I like. Now sometimes you're going to get an image that doesn't work. For example, here the dog isn't riding the bike and it looks a little bit distorted toward the back wheel there. So if you see things like this that aren't perfect, you can use that information to help Adobe train Firefly's models to make them better and better. If you hover over the large version of an image, you can click this thumbs down or thumbs up button. And if you click thumbs down and then tell us more, you have an opportunity to give feedback about the image that you thought wasn't quite right. And Adobe is also committed to generating content that's not biased or offensive. So if you see anything of that nature in a generated image, you can click this red report flag and give Adobe more info about that too. So that's a quick look at using text prompts to generate images in Firefly. Adobe is trying to make this as simple as possible for you, the creators, so that you don't have to become prompt engineers or experts in order to generate great images with Firefly. Another way Adobe's trying to make things as simple as possible is with the very exciting one click style inputs that are included in Firefly, and that's what we'll look at in the very next movie. So don't miss it.