Exploring the Creative Possibilities of Generative AI: Transforming a Photo with Artistic Styles and Imaginations

I’ve been doing some more experimentation with applying Generative AI to my existing images and thought I would share. I started with a base image of a photo of a Cardinal that I took this summer. Notice the watermark in the corner, that’s going to show up in a lot of the generated images and shows how heavily influenced they are by my original photo. As images go it’s fine, I was practicing with my telephoto lens and captured the cardinal in a pool of light, which is nice, but it’s nothing very exciting.

First off let’s see what happens if we just apply some common art styles to the photo. I used the same “seed” for all these prompts which reduces the randomness and makes for a closer comparison, so if you see a lot of common poses, that’s not a limitation of Midjourney, its deliberate on my part.

One in most of my prompts I do reinforce that the bird is a cardinal to avoid random other birds getting mixed into the images, one effect of that is that most of these images show a crest on the bird even though there isn’t one in my original photo. These are all pretty successful reinterpretations of the original image.

Then I thought it was time to get a little more adventurous and push the image further from its default state. For the most part these worked quite well too, although to my eyes the sand sculpture looks like it’s made out of clay with sand around the base. I believe that’s the result of mixing the two images together. The light paiting one is an interesting case too in that I first had to generate an image of some light painting and then combined that as well to reinforce the style I was looking for. For all these images I didn’t attempt to change the feel, just the look. Another time I may experiment with using emotions in the prompts as well.

The final four images are my absolute favorites. Here I stopped worrying about reality at all and this is where in my mind the greatest potential lies for this sort of thing. Yes, it will be able to generate a pixel perfect “photo” but, why bother, we can already do that. Here it’s meshing my photo, my ideas and it’s learning together to create something different.