I have been playing around with various XL models and noticed most of them would not generate uwabaki or require wrangling for inconsistent results. Could not find any uwabaki XL LoRA either. After getting over my amazement I decided to do the right thing and test the training funtionality of the site (because I have not trained a model before and this seemed like the most frictionless option).Turned out fine I think, so might as well share.How to:Use uwabaki to get uwabaki, add color soles and/or color accent to adjust the coloring.Used 48 images for training I believe, photos and drawn art, various aspect ratios and resolutions.v2:This time over 120 images were used, auto-tagged by Civitai and then all edited by me. Again a mixture of photos and drawings. Should be better than the 1st iteration, but no guarantees. Trigger words edited to hopefully make prompting more precise.

Files with SHA256: 8a5b7576da2afe40a8efe7e9960083c06aa9d34d63cef75a8d54f163dff8490e