Balanced and retrained with "Pony Realism v2.2 Main + VAE" on kohya_ss with "shuffle captions" and "randomized crop" on. Uses the same bad captions as v1.0, and may have been trained on images resized twice. The tag "inflatable" often makes everything colorful.
Balanced and retrained with "Pony Realism v2.2 Main + VAE" on kohya_ss with "shuffle captions" and "randomized crop" on. Uses the same bad captions as v1.0, and may have been trained on images resized twice. The tag "inflatable" often makes everything colorful.