Based on the works of wet tea.
v2.0 is trained on the images published in 2025
wettea-v2.safetensors