Hi guys, this is my SD XL LoRA trained with paintings from the argentinian artist Julio Le Parc
I´ve come to find that there are very few SD models and LoRAs for geometric abstractions so I decided to train my own. Geometric abstractions are quite hard for generative models to develope correctly because of the organic nature of noise diffusion and I´m still trying to get the best results out of this one. This is just the the first generation but I´m gonna keep updating the model to make it even better.
Julio Le Parc is a very famous argentinian painter that went through many different geometric abstraction styles along his carreer. The model is trained with almost all of his work, I only left out his very earliest naturalist paintings.
Hope you like my work and you can follow me on instagram to see the training process and all my other latin american artist trained SD models and LoRAs.
Description
Improved training algorithm and more epochs
FAQ
Comments (10)
Thank you, are you sure this is a LoRA 1.7GB?
Yes, with a huge dimension and low alpha. 256 or 128/1 probably.
@LDWorksDavid yes, it´s quite heavy because of that, geometric abstractions are hard for most diffusion models and I found that those parameters where the best to get good results.
cool, more geometry LORAS plz, these are amazing to have in workflows
Thanks, I'm working on another geometric LoRA with the artworks of rogelio polesello, will be publishing it soon
@nicolasmariar amazing friend! These ultra-connected hyper-continous lines, such a Franco Grignani lineart, .. its almost an art itself to imitate it. Its so insanely difficult to make perfect stable lineart.
Btw, if you find some more "extremely hard lineart" let us know . Rogelio posello, never heard of him, but it seems cool. Im looking at his stuff, looks very much like franco grignani, aka "retro lineart".
I wonder what other artists are "hard to imitate" ?
Make more LORAs please!
1.7g for a lora is outrageous!
How about a Flux LORA. This is my most used LORA for XL and it gives wonderful results, nothing quite like it. FLux, and Flux Kontext please!
Can you clarify the differences between 2.0 and 2.1?
I only discovered just now that there are 2 versions. I've been an avid fan of 2.1, and wondering if there is any point in trying 2.0 for a different look?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.