If the face isn't clear, try using "upper body" or "upper body focus" in the prompts. Generating the whole body tends to reduce the fidelity of the face.

When you have a lot of the same composition, it's easy to create a variety of compositions by combining dynamic angle, sitting, running, fighting stance, etc.

Setting the display to R/Z/ZZZ will hide some rollers and sample images : https://civitai.com/articles/31387/

I like Magia Record the most. I like it more than the original Madoka Magica series.

seaart : https://www.seaart.ai/ja/user/a0d65c884afd59ff6ac5d6787c0c4ffa?u_code=UZLW0XNF