I still also don't even know what I'm doing but it still obedience and didn't hijack prompt.

Recommended settings

- LoRA weight: 0.8 – 1.0

- Sampler: DPM++ 2M SDE, beta schedule

- Steps: 30, CFG: 4.0 (Anima Base)

- Dimension: 1:1, 16:9, 1:286:1

Notes

- Trained on a 75-image dataset; best results around weight 0.8–0.9.

- The model can produce NSFW output depending on prompt — mark your own generations accordingly.