Moss-Image | kouchan3 Style Lora

BASE Model:Moss-Image-Base-bf16

This is my first attempt at training a style LoRA for Moss. In actual inference, it isn't a 100% exact match—it captures about 80% of the original style.

PS: Adding other style prompts or artist tags will cause style drift, so please keep that in mind when prompting.

Generation & Settings

The showcase images were generated using DMD distillation, which you can also download and try if you're interested. Alternatively, if you prefer not to use distillation, you can use the following standard inference settings:

Sampler : Euler

Scheduler : Sgm_Uniform

Step: 35-50

CFG: 3

Base Model: Recommended to use the Base version for inference instead of the SFT version.

Krea 2 | kouchan3 Style Lora

BASE Model:Krea 2 Raw

Training on ModelsSope