SDXL model Training (XL 1.0) and Merging (self).
The sample is 512x768 doubled using R-ESRGAN 4x+ on Hires.fix.
Adjustments were made to ensure clean output even at low image resolutions.
hkxl_v2X.safetensors