In this new series, I’d like to periodically share some of my favorite AI art creations. I have recently taken a serious interest in testing the capabilities of available tools. And I must say, I am quite impressed with my early experiments. Being a writer, I appreciate the power of words to conjure images in a reader’s mind. But being able to convert text to image on a physical medium, electronic screens in this case, is a powerful feeling. Prompting the algorithm to create images has prompted me in turn to expand my imagination.
I would like to show you some of my most interesting creations from the past week not only out of joy at the aesthetic results but also to share how the algorithm interpreted words to generate what it did. I hope you enjoy.
Without further ado, here are 5 of my favorite images from the past week. Clicking on the images will take you the image posts on NightCafe, where it was generated. If you’re curious, it will also show you additional information about the models and prompts used. You can also check out my other creations on NightCafe by visiting my profile there.
Soaring
Composition: “colorful bird gliding across the morning sky”
Modifier: “in Gouache Style, Watercolor, Museum Epic Impressionist Maximalist Masterpiece, Thick Brush Strokes, Impasto Gouache, thick layers of gouache watercolors textured on Canvas, 8k Resolution, Matte Painting”
This was one of the first images I generated using AI. I was looking to depict simple subjects at first. I tried prompting for a flying or swooping bird but the algorithm struggled with depicting flight and correct anatomy. To make things easier for it, I prompted for a gliding bird. I have since become better at negative prompts (to prevent the algorithm from drawing wrong anatomy or other unwanted features).
Archway of the Moon
Composition: “Full Moon seen through a floral archway leading to a forked road”
Modifier: “cel shaded pop art”
I wanted to be a bit more adventurous with my composition and so I opted to be more specific. The algorithm doesn’t know how many moons you want in your image and what size they need to be. Remember, that these models are trained on real photographs and fantasy art so they can’t tell what’s supposed to be real. Getting it to also frame the moon within the ‘archway of flowers’ was good luck. Yes, there’s a fair bit of rolling the dice and hoping to get lucky in AI art generation.
Judicious use of prompts can make it manageable but you must realize that you won’t consistently get exactly what you were looking for. I didn’t get it to draw the forked road despite many attempts.
Reply to all
Composition: “(roman emperor: 1.3) wearing (gold wreath: 1.3) at his table in his imperial chamber (shocked expression: 1.3) looking at his laptop”
Modifier: “cel shaded pop art”
Generative AI art is, I believe, a genuine competitor to stock photography and art. Let’s suppose I was writing an article about challenges faced by the elderly in an age of rapid technological change. It would take quite some time to find a suitable representative image in a stock photography website. Commissioning work for one image could be expensive. But a prompt containing few words creates an evocative image. I appreciate that AI art unshackles the imagination. It allows blending of multiple ideas and styles that lead to interesting results.
Reflection
Composition: “woman’s reflection in an ornate water basin”
Modifier: “minimalism ink drawing”
As you saw in the ‘Reply to all’ prompt, I had by this point begun using weights in prompts. Through weights, you can tell the algorithm to prioritize (or de-prioritize through negative prompts) certain features, styles, or modifiers. When I created this image, I was still trying out different styles just to see how the algorithm interpreted them. The randomness of it all sometimes produces evocative results. It’s important, however, not to let AI dictate your thinking.
Parshuram
Composition:
“Ancient India Vedic sage with his axe. (1.5)”
“Vast, picturesque, diverse, wild, charming on the wide riverlands where the wind plays a melody, the violent storm carries its wild dance. Clouds black as pitch, lightning like flashes, the wind blows dances, it has no limits. (1.3)”
Modifier: “Alfons Mucha poster style, ultra detailed, ultra maximalist, HD, bright fire colors, epic cinematic, ultra quality, beautiful (0.8)”
One of my absolute favorite creations from the last week. It means to depict Parshuram, an avatar of Vishnu. As I noted in my first article on this website, large language models (and AI algorithms in general) are only as good as the data fed to them. Given the large amount of western themed literature and media on the internet relative to that from the rest of the world, a certain amount of bias will creep in to most results. Keeping this in mind, it is necessary to guide AI into depicting certain scenes instead of literally telling it to draw something specific (like Parshuram in this case). Another notable aspect of this image is the blend of styles that I had observed in two separate images from two different creators. Experimenting with styles and modifiers produces impressive effects. It is teaching me a lot about different art forms and techniques.
What did you think of these images? What are your thoughts on AI art? If you’d like to see more of my AI art, please visit my NightCafe profile and if you’d like to discuss more generally, please visit the Contact page on this website. I look forward to presenting another selection of images next week. Stay tuned!