Another one trained with Midjourney pictures
Trained on a complete new generated picture set.
I personally think it is better than V1.
BMAIW_sd15_scaled_128_v2.safetensors