Hey everyone! It's been a while! 😁

I'm excited to present my brand new project, "Crossed Nebula."
It was originally conceived for Illustious, but it took so much time to prepare that it ended up on Anima_preview3, which I think has a lot of potential!

v1: A dataset of just over 4000 images trained with Anima_Standalone_Trainer at 1344x1344 resolution.

I merged two saves from the same training environment for maximum stability.

A few usage tips:
- Quality prompt : masterpiece, best quality, score_9, score_8, score_7, highres,

- This model is really designed for the "flat" anime style.

But it might also surprise you in other styles (or not 😅).

- A minimum of 30 steps is required.

- cfg scale: between 4 and 5.

- Sampling method: Euler_Ancestral or, even better, er_sde.

You'll need a fairly robust scheduler: Sgm_uniform or beta57 (install the RES4LYF node in Comfyui for this).

Have fun! 😉