Exactly.AI
I’ve got to share some of these images that I’ve been making with Exactly.AI (Durer.AI)
I was able to train a model on my paintings using about 15 images + a language component (describing the art that I’m uploading for training).
Here are some generations from the first model I trained. It’s kind of fun to play with the prompts as a parameter in regard to generations from my own models. This is something you can do with current LLMs (Large Language Models), where the GANs (Generative Adversarial Networks) or CANs (Creative Adversarial Networks) have used no prompting.
I’ll show you more of why this could be fun, but first, this first set of images was prompted using words that were very close to the words that I input for training, making them closer to the actual style of what they were trained on.
This next set is what happens with the subject (above is abstract botanical), when the prompt changes from the original input.
Here, I switched out parts to add different subjects like palm trees, crows, gnarled trees, birch trees, coral, cacti, landscapes, skulls, and cats.
Artists can use this kind of thing as a study to see how one style might transfer to a different subject, or to explore different variations on their work before they set to it with paint.
One subject that this model is really good for is leaves - since so much of the work it was trained on was botanical abstraction. I loved how these came out. The prompts included much of my original text that described the style, material, and colors, plus the description of leaves falling and blowing in the wind.
I trained another model with some newer paintings, and here is what they look like:
Again, I kept the prompts close to the original prompt used when training, and that keeps the results looking like mine.
For me, using AI art tools contributes to a process of reflection, contemplation, and integration.
I’d be curious to hear how other artists who paint intuitively have found the process of training an AI model on one’s own work.
FEATURED POSTS
As part of the @thrivetogethernetwork, and the Taking. Up. Space. global grassroots initiative, #takingupspace2023 I’m seeking submissions for a collaborative project, set to debut in May. This project is called “Representations” and the intent is to educate and encourage womxn to take up space and affect the future at the intersection of art and technology. All of the collective work will be used to train an artificial intelligence model, which will be used to generate a new body of work reflecting the collective. I’ll be sharing the process of creation through live sessions and videos, and the final body of work will be available as free-to-collect NFTs.
Call for art ends: February 26th, 11:59pm EST