Welcome to Art Universe SDXL
The SDXL version is a big experiment in creating checkpoints for generating creative images.
v. 7.0 SDXL - A revision of the checkpoint. Models 4, 5, and 6 have been reworked, and the cfg generation has been increased to 6-7, up from the previous 4-5.
Friends. The idea for the implementation of Art Universe has come to an end. Perhaps a final checkpoint will be released with the best possible implementation of photorealism. Pay attention to Incredible World - this is a related model since they have the same basic checkpoint.
Special thanks for the opportunity to use LORA from Clumsy_Trainer.
This model was created by mixing several different models with each other, as well as adding about 10 LORA. This is an interesting model for creating images in the style of comics or cartoons.
The CFG scale must be at most 6 and the sampling steps must be at least 28. I recommend using hires and upscaler. I usually use Clip Skip 1.
"Bad-hands-5" and "EasyNegative" are great for embedding negativity.
I use ADetailer to restore faces.
This branch of models is a continuation of my work on UniverseStable models. These were the first models in which I did not adhere to a single style, there are also photorealistic models, magazine photography, artistic retouching, analog photography, a fantasy model, as well as two comic and animation models. I will be glad if you show interest in this work.
Description
FAQ
Comments (5)
Please, stop merging the LoRa I trained into your model, train your own or AT LEAST give credits. Training a model cost resources and time (also, I pay for the image used in my training) and if someone ask not to use the model as a merge you should respect that
I apologize, I did not carefully look at the restrictions. How do you want me to fix the situation. Deleting a model, or links to your LORA, your account?
@wildzzz a simple link pointing to the LoRa you incorporated at the end of the description would be more than enough. I would just ask to watch closely to those restrictions next time. Since your main goal is merging you should at least check if merging is allowed. I still have to thank you because you were sympathetic and responded very politely. Next time just ask and put a reference to the original network and it will be okay from my side.
@Clumsy_Trainer Great. I am grateful that you were so friendly about this situation. Your LORAs are interesting and unique.
@wildzzz Thanks, as I said, I put a lot of time and effort into my LoRas and test them extensively to be sure they are solid enough to be shared, that's why I'm so "protective". Glad you appreciate my work