Artificial Intelligence (i.e., AI) is the talk of the town. From coming up with summaries of long documents and data to explaining what a certain computer code does, there are many ways that AI has come into our lives. Though I won't be using an AI to write my blog posts any time soon (actually, never, since it's the writing I enjoy doing), I found AI that can create artwork in almost a minute or two quite amazing. Just by putting in some prompts, such as setting the style and content of the artwork, people see results of great quality so easily. When I was young, I dreamed of being able to magic a drawing based on my imagination since it would always turn out differently when I drew it out. AI is not magic, but I feel like it's almost as cool as that in some ways.
So, I tried out four different AI art engines and fed each of them four different prompts. You'll be able to see the differences (and perhaps some similarities) between the AIs. If the AI engine gave me more than one result, I chose the one that best suited what I imagined when I put the prompt in. The four engines I used are 'Midjourney', 'Dall.E 2', Adobe Firefly, and Stable Diffusion XL. I didn't choose them for specific reasons -- these were just what I had at hand.
Prompt 1: minimalist style concept art of a woman in a red swimsuit relaxing in a pool
(top left: Midjourney; top right: Dalle.E 2; bottom left: Adobe Firefly; bottom right: Stable Diffusion XL)
Prompt 2: a retro-style romance movie poster illustration
(top left: Midjourney; top right: Dalle.E 2; bottom left: Adobe Firefly; bottom right: Stable Diffusion XL)
Prompt 3: bird clock, 3D design
(top left: Midjourney; top right: Dalle.E 2; bottom left: Adobe Firefly; bottom right: Stable Diffusion XL)
Prompt 4: a realistic drawing of a 20-year-old Korean girl with long brown hair, brown eyes, and an oval face
(top left: Midjourney; top right: Dalle.E 2; bottom left: Adobe Firefly; bottom right: Stable Diffusion XL)
Wow! What a great experiment! I was really impressed with what Midjourney came up with -- it didn't take long to create the images, but the quality, as well as its ability to understand the prompt, seems to be the best of the four. I also liked the results of Stable Diffusion's work, but it was more technical than the rest, so I don't think I was able to use it to its full potential (and it took the longest to produce the artwork, sometimes coming up with errors). Dalle.E 2 seemed to understand the prompt, but the quality wasn't as intricate, while Adobe Firefly struggled with the forms of the figures in the images. Which AI drawings do you think turned out best? Which engine is the best to use?
I can't believe I live in a world where I can create a pretty good drawing by typing a few words in. Even if I don't use them as they are, and decide to make a few tweaks myself, it's amazing how much technology has developed. I feel like I can use this to create wallpapers for my phone the way I want it to or create concept designs before I start working on an artwork. But there's also the problem of real artists' work being underappreciated because of this new and easy way of creating artwork, as well as the issue of artists' artwork being unjustly taken into the AI's learning database without consent. We should step forward with ideas to prevent potential harm that can arise due to the nature of AI.