This model is a "fresh start" compared to others (agartha, eden and yomi).
It was developed starting by counterfeit and then improved with merged of my nirvana and eldorado private models style, then it was stabilized with few more merge and the result finally match my taste.
I had the need of a model that had no issue with prompt, and holy moly if i've cooked something. This model is still a clip skip 2 model, but has more flexibility than other models letting you mix a prompting style that uses danbooru tags with a more straight forward prompting method like the clip one.
As you can see from the images in the gallery this model excells in 2d art with flattish colors, but it can also be used with some prompt and lora to generate more bright and illustration like images.
If i had to find a flaw in this model it would surely the hands, i couldn't make it better than other models, sadly it happened kinda often that i had to generate an image using the same seed with some lil variation to fix an awfull hand.
It works good (like any other model) with euler a, at clip skip 2 and cfg of 7 (if you use more tokens than 75 for your prompt, you might need to increase it).
It's strongly suggested the use of highres fix and of an upscaler (i personally use 0.2 denoise strenght and 10 steps with animesharp upscaler).
Adetailer works aswell, but it's not mandatory.
As always i hope you like this, and if you do you can also supporting me by sharing your images, reacting to posts, leaving a review or even by tipping buzzs.
This model is a "fresh start" compared to others (agartha, eden and yomi).
It was developed starting by counterfeit and then improved with merged of my nirvana and eldorado private models style, then it was stabilized with few more merge and the result finally match my taste.
I had the need of a model that had no issue with prompt, and holy moly if i've cooked something. This model is still a clip skip 2 model, but has more flexibility than other models letting you mix a prompting style that uses danbooru tags with a more straight forward prompting method like the clip one.
As you can see from the images in the gallery this model excells in 2d art with flattish colors, but it can also be used with some prompt and lora to generate more bright and illustration like images.
If i had to find a flaw in this model it would surely the hands, i couldn't make it better than other models, sadly it happened kinda often that i had to generate an image using the same seed with some lil variation to fix an awfull hand.
It works good (like any other model) with euler a, at clip skip 2 and cfg of 7 (if you use more tokens than 75 for your prompt, you might need to increase it).
It's strongly suggested the use of highres fix and of an upscaler (i personally use 0.2 denoise strenght and 10 steps with animesharp upscaler).
Adetailer works aswell, but it's not mandatory.
As always i hope you like this, and if you do you can also supporting me by sharing your images, reacting to posts, leaving a review or even by tipping buzzs.