Use Euler/Normal.
It works way better in horizontal since there were only horizontal images in the training data.
wowcine_lora.safetensors
