---
license: openrail
---
This is an attempt to "unmerge" the sd1.5 base from the AOM3(a1b) model, and thus hopefully improve the quality.
This is a new experiment, shared to improve research on this.
For best results, you should probably use steps=30+