AI Text Prompt UnBundling Techniques
Today, my experiments will have us creating AI Text Prompts with “Unbundling” techniques. Unbundling techniques caught my eye as a possible solution to generating digital art while being respectful of living & recently deceased artists, their families, friends and fans: by NOT putting the human artists’ names in our AI Text Prompts. Adding a human artist’s name into our AI Text Prompts for Art Generation quickly buries / erases those artists names off all search platforms and replaces them with AI Generations that are also being mislabeled with another person’s name as the ‘artist’.
We can avoid harmful practices with Unbundling and similar techniques that are respectful, authentic and do no harm while helping humans to learn, understand, play and experiment. I will show how AI can assist us with learning more about Art History, Mediums and Techniques when we are not yet sure of how to describe art composition, materials and techniques.
The benefits of using the Unbundling technique are described by Bakz T. Future as “you get results which are slightly different and appear distinct, maybe even a little bit new, but also subconsciously familiar to your audience.”
Unbundling is described by Bakz T. Future as a technique for DALLE-2, but of course it can be used with any AI Generator too. Here is a direct link to his article ‘DALL-E 2 Unbundling‘ where he describes how to do this technique by using GPT-3 to write your prompt and how to write the prompt yourself without AI assistance.
EXPERIMENT 1: Cat
I think I want a cat painted in the style of The Creation of Adam by Michelangelo. These experiments will help feed curiosity about Art History, Techniques and Mediums regarding Michelangelo.
I did a quick google search for “AI Text Prompt, Unbundling technique”. After a few quick reads, I created the following two prompts and got the following results from the Stable Diffusion python script I run on my own pc laptop. I also used the same text prompts with DALLE-2. You can see the results in the rest of this article.
AI Text Prompt: local Stable Diffusion
“fresco painting of a cat on a sunny day, in the style of The Creation of Adam by Michelangelo, digital art”
AI Text Prompt: DALLE-2
“fresco painting of a cat on a sunny day, in the style of The Creation of Adam by Michelangelo, digital art”
AI Text Prompt with Unbundling: local Stable Diffusion
“fresco painting of a cat on a sunny day. The painting is highly symmetrical and dynamic with a strong diagonal line in the composition, to create a sense of tension and movement. The color palette is earth tones, browns, greens, and grays. Sfumato blends colors together to create a soft, hazy effect. Digital art”
AI Text Prompt with Unbundling: local Stable Diffusion
“fresco painting of a cat on a sunny day. The painting is highly symmetrical and dynamic with a strong diagonal line in the composition, to create a sense of tension and movement. The color palette is earth tones, browns, greens, and grays. Sfumato blends colors together to create a soft, hazy effect. Digital art”
Experiment 1 cats do not look like the style of ‘The Creation of Adam by Michelangelo’ in the way I was expecting. Let’s figure out what we can do differently with that Unbundling technique. Let’s figure out how and if the ‘Sfumato’ technique is even in that result. Off I went to look at as many images as I could find on the internet that show examples of Sfumato techniques.
I searched Google Images for “michelangelo“, “sfumato, michelangelo” and “sfumato” and can see that the colors, blending and art materials are certainly looking like sfumato. The difference from my expectations and the Unbundled result are that the cat’s body & face shape composition are not photoreal nor realistic. Maybe we can add some parameters to define realism such as: ‘high detail, high definition, high quality, photo-realistic, ultra-realistic, hyper-realistic, 8k’ as described on our post Text Prompt Tips: diy
ChatGPT
Learning is a lifelong adventure for me, and I am currently learning about Art History and Techniques. ChatGPT helped me figure out how to properly describe the techniques and mediums that I like about the art I’ve seen from Michelangelo. ChatGPT provides me with many paragraphs for a description and I use that as a reference to look up more information and that process helps me write my own AI Text Prompt as seen earlier in this article. My first request from ChatGTP was to help me gain more understanding about the art style, characteristics and techniques that Michelangelo used when creating ‘The Creation of Adam’. I then did more searches and reading about the info that ChatGPT shared with me.
I had never heard of ‘sfumato’ before ChatGTP mentioned it to me, so of course I had to find out more about that with searches, reading and asking ChatGPT for more info.
Now, I have more interesting words and descriptions to go searching, reading, learning and experimenting with! What is ‘Chiaroscuro’ and how can I use that in my AI Text Prompts to get the results I hope for?
Aha! I need to request my new AI Art Generations show Chiaroscuro techniques. Now, I can either ask ChatGTP for a more specific AI Text Prompt or I can write it myself with this new information. I’m going to go look through Google Images and the many Art platforms online to see what human beings have made and are making with these techniques using these ‘new to me’ words as a search term.
EXPERIMENT 2: Cat
AI Text Prompt: local stable diffusion
“fresco painting of a cat on a sunny day, in the style of The Creation of Adam by Michelangelo, high detail, high definition, high quality, photo-realistic, 8k”
AI Text Prompt: DALLE-2
“fresco painting of a cat on a sunny day, in the style of The Creation of Adam by Michelangelo, high detail, high definition, high quality, photo-realistic, 8k”
AI Text Prompt with Unbundling: local Stable Diffusion
“photorealistic fresco painting of a cat, chiaroscuro, sfumato, depth, realism, light, shadow, movement, depth. hazy, atmospheric. cat face illuminated by a beam of light, contrast between light and dark, high detail, high definition, high quality, photo-realistic, 8k”
AI Text Prompt with Unbundling: DALLE-2
“photorealistic fresco painting of a cat, chiaroscuro, sfumato, depth, realism, light, shadow, movement, depth. hazy, atmospheric. cat face illuminated by a beam of light, contrast between light and dark, high detail, high definition, high quality, photo-realistic, 8k”
I’m happy with how the second Unbundling text prompt (local stable diffusion) experiment turned out. Now that I see the text prompt results, I have more understanding about another aspect of historical fresco paintings using chiaroscuro and sfumato techniques that I enjoy; and that is the inclusion of more than one character in a shared group activity, environment details and that ‘beam of light from god’ on the face of the main character. The results for experiment two are nice enough that I will most likely use variations of this text prompt with the unbundling techniques in future projects.
As for the standard AI Text Prompt result for experiment two, it does show typical monstrous appendage results. This unedited monstrosity can be used as an Artist’s graphic design exercise for correcting appendages with digital graphic design software. I’ve seen creative artists are taking still photos of their own hands in the relevant angle and lighting, then pasting the digital image of their own hands, feet, neck into the AI Art Generated images. Results that are not desirable as high-resolution art when viewers look at the image closely are still useable and even desirable in some art projects such as mobile video games, 3D environments, virtual reality where much lower resolution digital images are desirable and not expected to be viewed up close in high resolution (mobile devices have limited space and low resolution helps them perform better). Some of the AI Art Generated images in this post’s experiments can also be used as 3D shading, texture, material, tiling projects where close up details are not going to be seen while the general overall image will be seen in a much larger 3D environment (wall mural, floor tile / rug, etc).
EXPERIMENT 3: Cat
This experiment will build on what I noticed with the results of our previous experiments. I want results that show more of the group activities, dynamic movements and detailed environment that I now understand are the elements of Michelangelo’s work that I admire along with the techniques of chiaroscuro and sfumato. My previous ChatGPT results will help me figure out how to describe my expectations another way.
AI Text Prompt: local Stable Diffusion
“fresco painting of a family on a picnic with a cat, in the style of The Creation of Adam by Michelangelo, high detail, high definition, high quality, photo-realistic, 8k”
AI Text Prompt: DALLE-2
“fresco painting of a family on a picnic with a cat, in the style of The Creation of Adam by Michelangelo, high detail, high definition, high quality, photo-realistic, 8k”
AI Text Prompt with Unbundling: local Stable Diffusion
“photorealistic fresco painting of a family on a picnic with a cat, light, shadow, movement, depth, chiaroscuro, sfumato, depth, realism, light, shadow, movement, depth. hazy, atmospheric. cat face illuminated, high contrast, high detail, high definition, high quality, photo-realistic, 8k”
AI Text Prompt with Unbundling: DALLE-2
“photorealistic fresco painting of a family on a picnic with a cat, light, shadow, movement, depth, chiaroscuro, sfumato, depth, realism, light, shadow, movement, depth. hazy, atmospheric. cat face illuminated, high contrast, high detail, high definition, high quality, photo-realistic, 8k”
Experiment 2 with Local Stable Diffusion Unbundling is my personal favorite in this series of experiments today. I also like Experiment 3 Unbundling with DALLE-2 alot but my prompt creation was not relevant enough to get the expected results. I know today’s experiments will have a strong influence on my future art production pipelines while minimizing and removing harmful tech industry practices that many humans continue to call attention to for many years. We can all do better with our learning, experimentation and practices and it looks like this Unbundling Technique is a great start to being respectful of each other and ourselves.
Today’s experiments helped me to learn what I find appealing about ‘The Creation of Adam’ by Michelangelo and that led me to learn about Chiaroscuro and Sfumato techqniques. I also learned how light, shadow, blending is used to create movement and dynamic scenes that are realistic even during ancient times when Octane Renders were not a thing yet. Now, I’m off to find out how those materials in Michelangelo’s frescoes were made, what is the recipe? What plants were used? How can I recreate sfumato and chiachiaroscuro techniques in Krita? So many questions, curiosities and experiments for future posts here!
Do these experiments inspire you to go figure out how to use your own human hands to play with sfumato and chiaroscuro techniques away from the computers? Are you running your own AI Art Generation experiments? Is your content and experiments safe for all ages? If yes and you would like your experiments to be featured on this website with a link to your experiments and social media: Contact Us. If your experiments look like something our readers will be interested in, we can create a friendly for all ages post that features your experiments.
If you like what you see on this website, please consider showing support by liking, subscribing and sharing our posts on social media, to see our newest AI Text Prompt Ideas, Experiments, Failures and Successes.
