An SDXL checkpoint created through training on SDXL 1.0 and merging my own models.
Sample settings
The sample images were generated at 512 × 768 pixels and upscaled by 2× with Hires.fix using R-ESRGAN 4x+.
Description
FAQ
Comments (2)
so is meant to be an anime style model ?
Since it outputs coherent images in 512, does that mean the training was done at 1.5 levels or resolution?