I created this LoRA as a bounty entry but it wasn't picked up, so I decided to share it. Have fun playing with it !
It has been trained on the default Illustrious checkpoint on CivitAI, with various pictures (females only) in the Soul Land style, with high resolution (1.4M+ pixels) and a good quality. The dataset inclues mostly asian style outfits, but also modern clothes and backgrounds.
It can be used alone or along other character and/or clothing LoRAs. All example pictures are "simple" outputs, which means they only went through a face detailer. No other post-processing has been made (upscale, manual edit...)
The WAI-NSFW-illustrious checkpoint gives overall good results. But small 2D style can remain with this checkpoint.
Description
FAQ
Comments (3)
A dataset can't really be called "diverse" if its training subjects absolutely weren't, that limits its utility and range of usable application. Just n.b.
In some languages, the word diverse simply means 'an assortment of', not 'of substantial variability' as it does in current English usage. Using a comma rather than a full stop as a decimal delineator makes me suspect the OP uses English as a second language, just like myself. Your point is still valid, but if you suspected some sort of intent at diluting the word, I am pretty sure it wasn't there.
Yeah, I'm French. Nobody's perfect :D
As stated on Wordreference, diverse is a synonym of various. But it's probably not the most used word in this context, I must admit. What I wanted to mean is, Soul Land takes place in a fantasy asian world, where people wear traditional clothes. So I added some modern clothes (always in the same drawing style) in order to my LoRA draw them easier if needed. So it's various not in the drawing style, but in the clothes type.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.