From the course: Adobe Firefly Essential Training

Understanding the difference between Firefly models - Firefly Tutorial

From the course: Adobe Firefly Essential Training

Start my 1-month free trial

Understanding the difference between Firefly models

“

- [Instructor] Adobe Firefly is driven by a large data model, and that algorithm is giving you the results that you currently see onscreen. So this really big algorithm is driving a lot of the functionality, not just here within the Adobe Firefly web page where we're currently using the Text to Image functionality, but also when you launch Photoshop and you're using additional functionalities, like Generative Fill or Generative Expand, or in Illustrator when you're using Text to Vector, or Adobe Express or any other application that's fueled by Adobe Firefly. So what Adobe's doing is, it is regularly fine-tuning and updating that specific algorithm to make it smarter and to make sure that you are getting far better results than what you were able to do before. And this results not just in having a better quality, as in a visual quality image or an image result, which might result in larger resolution, as in more pixels, or an image that's clearer or sharper or that looks less AI-generated or less CG-generated, for example, but this also allows you to interpret the prompts in a different way. Which means that Firefly is going to get smarter, and it's going to get a lot better at interpreting the prompt and interpreting the words that you type in here into this text box. Now, if you look at the model options on the left here of my screen, you can see that there's a Firefly Image 3 model, and there is a Firefly Image 2 model. And this is what I currently see at the time of this recording. So at the time of this recording, for me, Adobe has just released Firefly Image 3 as a preview version, and this is the version that comes after Firefly Image 2, which is released roughly in the month of October in the year 2023. So which means that, as you're using Firefly Image 3, which is the default today, you will get very specific results. But that does not take away that, maybe, in some circumstances, if you go to Firefly Image 2 and you switch models, you will get a different result of the same image. It's using an older version of the algorithm, but old doesn't quite mean obsolete all the time. So if you click Firefly Image 2 instead, you can even see that I now have photo settings available, which allows me to set specific settings, like aperture, shutter speed, and field of view, something we'll cover in a future video. And this is not available if you go back to model three, where I'm assuming, the interpretation of specific photo settings is already embedded in your basic prompt functionality. But if I go back here and if I go back to the Firefly Image 2 and then quickly click the Generate option, it's going to reinterpret the exact same prompt I have here in the prompt command box, but it's going to be using Firefly Image model two to give you the result that you're looking for. So let's look at the result here. Now, I'm getting a very similar result. It's taken in the same circumstances of the image. I've got four very smart dogs reading books by candlelight with a library of books in the background. But I can't help but feel if I go quickly back to the previous version by clicking the back button here, that I am actually seeing a more photorealistic result using Image model three. So depending on what it is you're looking for specifically, for example, something that has to look photorealistic versus something that might look a bit more CG-generated or that might look a bit more like an art piece, you might still want to consider going back to version two and look at that different interpretation of the same prompt. I've actually had a few cases where I did notice that Firefly Image 2 gave me a slightly better result compared to Firefly Image 3. But it all depends on the prompt, the settings you set, and your own expectations. What is it that you're visually looking for when you are trying to generate these images? So by default, keep in mind that there is Firefly Image 3, which is something that you will always be using throughout this course, but you also have a way to go back to version two. And just be curious and experiment and try and see if you can get a slightly better result using model two instead of three.

Contents