All licensing conditions for the Flux Dev model apply. You can find the details here.
Description
FAQ
Comments (11)
looks good, i'll give it a try later
#AwesomeSauce
Excellent Lora. Thank you for your effort.
V2 strengthens and improves V1.
However, I don't see the connection with the style of the 2 illustrators mentioned.
It reminds me more of the style of another artist but I can't put my finger on it, but I like it as is.
Yes, you are absolutely right.
As for Q Hayashida, I didn't try to recover his graphic touch but the underground universe of her Dorohedoro manga. My training base was largely made up of scenes from this manga but restyled in I2I with a graphic touch from Sergio Toppi (with one of my XL models).
Already the XL models reproduce the Toppi graphics quite poorly, but when you switch to Flux, it gets worse. Flux is a very rigid model that makes it difficult to express artistic styles. Thank you for your feedback
@Adel_AI ok, I understand your approach and it makes me curious. When you go from XL to Flux, is it in image to image or do you use a workflow that uses the scenes of the first artist by applying the style of the other to create the lora?
The style transfer in Flux is... pfff lol
So, I say to myself: ok, style transfer under XL (which is not so bad) then img2img + llm to have the quality of Flux.
It's a hell of a job.
Thank you very much for sharing ! Amazing LoRA, it improves on all aspects of Flux comics/illustration generation. Even when prompting for images not with the two styles it was trained on.
Thanks for the feedback
Is it possible to build the same for SDXL 1.0 ?
It's hard to work on multiple models and I have little time, I prefer to focus on Flux
Ok thank you for the reponse. Great model.