Marshmallow Fluff SDXL

A LoRA for Stable Diffusion XL trained on the character marshmallow fluff, a furry female character.

(Some sample images are at epoch 9. It performs way better in published final model.)

Note: Sometimes watermarks/text might appear, try using a negative prompt against it.

Usage

Add the LoRA to your prompt with your preferred weight (recommended: 0.7–1.0):

marshmallow fluff, <lora:marshmallow_fluff_sdxl:0.8>

Trigger words

marshmallow fluff

Recommended settings

• Base model: Stable Diffusion XL 1.0

• Resolution: 1024x1024

• Sampler: DPM++ 2M Karras

• CFG scale: 7–7.5

• Steps: 25–30

Training details

Training method: LoRA

Optimizer: AdamW 8-bit