This is my first attempt at working with MOSS Image SFT. This checkpoint includes the 8 step DMD turbo LoRA from lodestones.

It also includes a LoRA, trained on Civitai, using output from LexiVision-Z to transfer its style to MOSS Image SFT. By way of comparison, image #3 is the original SFT generation (with turbo LoRA).

Original Apache 2.0 license applies. Give it a go, post your results here.

*Edit: I have noticed a bit of "AI same-face" that I'll try work on for V1.1.