A legendary actress, I trained the network on photos from the 80s and 90s. Best result, use parameter 0.8-0.9.
Better photo preparation. The learning algorithm has been changed. Thanks to user maxhitman for providing the photo.
viver_v2.safetensors