Anima · Forge-compatible ZeroGPU

DiffSynth inference with Forge-style weighted conditioning and reversible LoRA/LoKr injection. The model is loaded once; each request's LoRA hooks are removed after generation.

Prompt

Negative prompt

Width

512 1536

Height

512 1536

Steps

10 60

CFG

1 8

Sigma shift (0 = default)

0 8

Seed (-1 = random)

Output（轻量预览）

页面预览可能被 Hugging Face 压缩；生成完成后请使用上方按钮下载服务器保存的无损 PNG。