In my previous blog, Testing Out the Beta Version of the New Firefly Integrated Adobe Photoshop Part 2, I explored adding images using the additive image generation feature and starting with a blank canvas to make something out of nothing using the text-to-image feature. With the right prompts, these features can save countless headaches attempting to find that right stock photo. What about the example of changing one's outfit by using the generative fill feature? Or what about creating an entire scene for a book with a simple description?
Changing Someone’s Outfit
Starting with replacing someone’s outfit. Seemed like a pretty simple concept. I had already replaced the background on the pictures. How hard could this be? To start this process, I wanted a picture where you could already see either the entire outfit or at the very least, most of the outfit. Once again using a picture of my fiancé, Erik, I started by using the quick selection tool to select just his clothing. I then prompted the program to replace the clothing with a button-up purple plaid shirt with dark blue jeans. I was shocked. Not the most realistic results but by far the best results after just one attempt.
Next, I wanted to see if I could replicate the results by selecting one clothing item at a time. This turned out to be a much more frustrating process. First, I selected his jacket and prompted the program to replace it with a black leather jacket. That part was easy enough. Swapping out the shirt proved to be a challenge. At first, I attempted to swap it for a plain black T-shirt. The problem was every time I clicked the generate button, it gave Erik a very odd neckline. Instead, I opted to swap it out for a plaid shirt instead. I hypothesized that since a white collared shirt and a plaid shirt have the same neckline; it would look more natural. My theory was correct. The last part of the experiment was to swap out his pants. Just like the jacket, this proved to be a straightforward process. I selected his pants and prompted the program to replace pants with dark blue jeans. You can review the results below.
Combining Two Pictures
I stumbled upon this trick by accident when I was scrolling aimlessly through Instagram one night. The concept of combining two pictures is not a foreign concept to those familiar with Photoshop. However, combining two pictures in a matter of minutes is always the real challenge. I started with something simple. Starting with a fresh canvas, I added a picture to it moved it to the left, duplicated the picture moved it to the right, and then selected the middle using the rectangle marquee tool.
Then, I prompted the program to combine the two pictures. Easy peasy. But I knew it wasn’t always going to be that easy.
Starting with two pictures that had similar backgrounds and two different subjects, I moved on to the next experiment. Again, starting with a fresh canvas, I dragged both pictures into the document. One on the left side and one on the right. Instead of selecting the middle of the two pictures and prompting the program to combine them, I thought I would attempt a different approach. I selected the entire canvas and then prompted the program to combine them. This turned out to be a horrible mistake. I shall spare you those results. Using my favorite tool, command + z, I undid the mistake. This time, I only selected the middle and again prompted the program to combine the two pictures. Is it the perfect example, no. However, it is a good start in learning what the AI software is capable of. You can view the results below.
Creating Something Out of Nothing
I have been looking forward to trying this feature out since I first started playing around with the online version of Adobe Firefly. I was curious to see if the Photoshop version held up with the online version of Adobe Firefly. I am pleased to say Photoshop indeed held up.
As an avid reader, I would love to see my favorite scenes come to life without spending countless hours illustrating one small part of a scene. For this experiment, between Google and brainstorming with my fiancé, I was able to put together successful prompts to input into the program. Starting with a fresh canvas, use the rectangular marquee tool and select the entire canvas.
For the first example, like always, I started with something simple, I prompted the program to create a dark and creepy forest at night with a tall, huge, hairy, and scary monster roaming about.
This next example proved to be the most challenging out of all the prompts. The goal I had in mind was to create a character you might find in a children’s book. As I have stated many times before, the program doesn’t do the best with faces hence why this one took a couple of tries to get just right. The final prompt that produced the successful result was: a cartoon ice princess with big round blue eyes, red rosy cheeks, soft pink lips, and light golden hair with a golden tiara, the princess is wearing long flowy blue and white robes with accents of gold, in the background she is on a snowy mountain with trees.
Saving the best for last, the final example was by far my favorite. Not only did I manage to get what I was looking for on the first attempt, it was exactly what I was thinking in my head. The prompt was as follows: in the style of a mystical painting, an orange, black, and white fox sits at a table in front of a glowing translucent crystal ball ready to predict the future.
If you are in any creative field, I highly recommend exploring this new software. The examples I gave for the experiments I tested out for the past three blogs, are just scratching the surface. While it can be frustrating at times to produce successful results, it is a necessary skill to acquire. People are already using AI tools like ChatGPT for everyday tasks and it will not be long before designers like myself are using AI consistently in all our projects.