No trigger word needed for this lora.
v2.1: Is closer to the artist style and coloring that v2 while maintaining good quality. I think I need to remove some of the image from the artist to actually get the style I want, since the quality of his works is very uneven.
v2: v2 is far from perfect, but better in every single way than v1. It still hasn't learned this style perfectly and focused a bit too much on background instead of characters.
Description
Lora trained without triggerword.
Lora trained on 80 range murata images at 768x, using kohya-ss/sd-scripts. The lora has been trained for 2.5k steps with TALON optimizer 6e-04 LR cosine batch size 4.
It is trained on Chroma-DC-2K, which I feel is a better model than HD.
FAQ
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.