Models were updated, and the merge method was changed, should have better quality overall.
realMoon_v80.safetensors
