If you explore it, it will be able to provide you tools to research how AIs generate image. Do they have a consciousness about their creations? If not, why they are able to place
so many references to uncousncious mind? Give it a go and see it yourself. Althought it is experimental and it's not 100%, read more on that below, please.
All images generated used whether Euler or Euler A with 20 to 25 steps CFG from 5 to 7. Increasing CFG and/or steps too much, makes the base model create undesireable results.
Although as in any image generation model, it all comes to hardware. Adjust those numbers according to those you usually set.
Note: The Artist refers to an idea on how image models "see themselves," that if they do see anything at all, of course. Anyway, as "the Artist" they render have any shape, explore it.
Model description
This is an Abstract|Style|Expimental image lora model, trained on
Here is a JSON representation of the model's metadata, including the information and citation.
{
model_name: The Arstit | Style for Illustrioux XL,
description: A LoRA model that infuses digital artwork with vibrant and imaginative styles, transforming images into captivating masterpieces,
trained_by: Robb-0,
license: Creative Commons Attribution 4.0 International (CC BY 4.0),
capabilities: [
Adds vibrant rainbow gradients to create eye-catching visuals.,
Create aristic Apple-esque emoji and avatars by adding 'circle' token in prompt,
Enhances special effects in digital illustrations, bringing subjects to (another) life,
Introduces fantasy elements, such as stars and galaxies, for a magical touch,
Add a new touch of art to Illustrious just to facilitate the process,
Captures intricate details, especially in textures and highlights;
]
}
Model Card:
The Artist - Illustrious Style
Name: The Arstit | Style for Illustrioux XL
Description: Unveil the artistic prowess of The Arstist Illustrious Style, a LoRA model designed to elevate digital artwork to new heights of creativity.
This model is a gateway to a realm where vibrant colors and imaginative themes intertwine, transforming ordinary images into captivating illustrations.
Key Features:
Rainbow Brilliance: Watch as ordinary hues burst into vibrant rainbows, adding a unique and captivating flair to your creations.
Digital Realism: Witness the model's exceptional ability to enhance the realism of digital subjects, making the unreal appear tangibly real.
Fantasy Infusion: Prepare for a magical touch, as stars and galaxies find their way into your artwork, creating otherworldly scenes.
Detail Mastery: Experience the precision in capturing intricate details, from delicate textures to subtle highlights, elevating the overall visual appeal.
Training: It's a-me Robb-0 who trained this model, I used some images I had gathered about space-cowboy-esque Apple-avatars and crossed with similar images which represented
artistically weird things like inside Stable Diffusion, the latent space according to Diffusers. Which means images generated by SDXL and Flux with the subject being the latent space and stuff.
Usage: The Arstit | Style for Illustrioux XL is a creative companion for
artists seeking to infuse their digital creations with a burst of colour, life, and fantasy.
Whether you're crafting book covers, game art, or simply exploring your imagination, this model is your gateway to a world of vibrant, enchanting visuals.
Aid with artistic / conceptual / abstract representations of AI depicted by itself. No image is the accurate figure of any AI, yet it may be useful to be paired with
conceptual texts, educative brochures, narratives, etc.
Can I use to provide an abstract image reference for AI in the article or book about AI I'm writing?
Yes, sure thing. I hope it works well and suits your need.
Attribution and License: This model was trained by Robb-0, licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0).
When using this model, it's not necessary but it would be nice a proper attribution is given to the creator.
As seen on samples, results not necessarily render as expected. Still they keep the style in the composition. Happens that there are at least concepts here - all intertwinted:
*Note: Sd 1.5, SDXL and Flux are able to render such types of images WITHOUT any LoRA model, but you can use those prompts above as inpiration.
May help you create Apple-like avatars
May aid in creating some abstract/artistic representation of how AI render inputs such as "latent space," "the artist," and some nice special effects.
That sort of "turbine" seems to be how image models represent Kalman Filters or Kernels. What I find it amazing is that no only this LoRA model, but many image models
tend to represent them like that. Maybe base models were pre-trained with such images?
Languages suggested are those which IllustriousXL and NoobAI XL seem to work best.
Compatibilities
The The Artist / Style image lora model was trained on Indigo furry Mix XL Noob AI EPS III.
Reason: It manages well both Dan/Gel-booru tags as well as natural language.
NoobAI EPS III model and finetunes: should be able to create nice images with this LoRA models (I tested and it works, samples pinned above).
I also have tested it on Illustrious 1, 1.1, 2 and 2 PRO. Results are dilluted. It's a hit or miss. Although they are more recent than EPS III, which may explain it.
Previous models such as the base Illustrious V0.1 STILL should be able to create images as in here. You may or may not need to increase CFG and increase the steps - or switch sampler to SDE++ (loads of artifacts)
Pony: It's not impossible that Illustrious Loras run on Pony. Especially when they are overfitting. This one does NOT work with Pony when I tested it. And chances are that Pony is able to render these things without any LoRA.
SDXL base: It's likely not to work. I have not tested it either, yet odds are it won't work.
WARNING
Since this is IllustriousXL, the lora has only cute illustrations and references, but base model can go easily from PG to XXX. I am not responsible for such images in any way. Please avoid and or do NOT use this
LoRa if you are underage, if you cannot answer for your own deeds, and if you live in a place or country which forbids everything. You agree that the responsiblity is yours.
DISCLAIMER
This is just a "small" and "experimental" image LoRA model, it's not infallible, although when it works it's cool - and I hope you have read the WARNING above, as those things are beyond my control. I may or may not keep working and developing the idea.
Download model
Weights for this model are available in Safetensors format.
This is an image LoRA model, you need access to a UI which runs Illustrious or NoobAI models and a slot for image lora. ComfyUI and Automatic1111 (among others) to run locally.
OR
Simply visit Civitai and the Lora is available for use on site as well: https://civitai.com/models/1370366
