First Release










Works slightly better with semi-realistic checkpoints (using CLIP skip 2) but it ultimately depends on the prompt. Has a tendency to go and will pull heavily on the training images if you don't specify the background and/or outfit details. It is flexible enough depending on the checkpoint to generate good results with proper weights and prompts.
Recommended weight from my tests: between 0.9 and all the way up to 1, also a low CFG scale of betwee 4.5 and 5.5).
My sample images are a mix of different checkpoints and samplers (Euler A or DPM++ 2M SDE Karras for the more illustration leaning ones, plain DPM++ 2M Karras or UniPC for the more photorealistic ones). CFG scale around 4-5, If the checkpoint was capable (and most current SD1.5 should be) I ran the base images on a resolution of 768x1024 with a HighRes-Fix of 1.5 (and ADetailer on occasion).
Other settings may work better or worse - feel free to share yours!
First Release
1. Hak untuk model yang diposting ulang adalah milik pembuat aslinya.
2. Pencipta asli yang ingin mengklaim model harap hubungi staf SeaArt AI melalui saluran resmi. Klik untuk mengklaim
