This LoRA is my attempt to recreate the style of Wenjun Lin, known for vibrant, dreamy, painterly fantasy art with soft colors.
Description
trained on NAIXL_vPred10 and should work better with vPred checkpoint. If the anatomy starts looking odd (hand issues, yeah), try lowering the LoRA weight. Personally, I find a weight of 0.6–0.8 to be optimal.
FAQ
Comments (3)
It is strange, the lora does not work for me. I enter the keyword and it does nothing. Also, I remove the keyword and although I enter a strength of 0.7 or 0.8, it does not activate the style.
Hi, this is a bit confusing. Are you using NAI? I was testing LoRA on noobaiXLNAIXL_vPred10 and noobaiXLNAIXL_epsilonPred11. I also tried generating on-site using NAI vPred, and LoRA definitely affects the style (you can check out my latest posted image, the red oni girl). The keyword doesn’t really seem necessary — I don't usually use it, though it was used in training.
If you are using A1111 or anything derived from that, be sure when the lora is entered in the prompt that the file name is the same as what the file is actually called. Had the same issue with some other models
Details
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.