metadata

license: mit
base_model: stabilityai/stable-diffusion-xl-base-1.0
tags:
  - stable-diffusion
  - stable-diffusion-diffusers
  - text-to-image
  - diffusers
  - lora
inference: true

sdxl-wrong-lora

A LoRA for SDXL 1.0 Base which improves output image quality after loading it and using wrong as a negative prompt during inference.

Benefits of using this LoRA:

Higher color saturation and vibrance
Higher detail in textures/fabrics
Higher sharpness for blurry/background objects
Less likely to have random artifacts
Appears to allow the model to follow the input prompt with a more expected behavior

Usage

The LoRA can be loaded using load_lora_weights like any other LoRA in diffusers:

import torch
from diffusers import DiffusionPipeline, AutoencoderKL

vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
base = DiffusionPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    vae=vae,
    torch_dtype=torch.float16,
    variant="fp16",
    use_safetensors=True
)

base.load_lora_weights("minimaxir/sdxl-wrong-lora")

_ = base.to("cuda")

During inference, use wrong as the sole negative prompt.

Examples

Left is the base model output (no LoRA) + refiner, right is base + LoRA and refiner. The generations use the same seed.

Methodology

The methodology for creating this LoRA is similar to my wrong SD 2.0 textual inversion embedding, except trained as a LoRA since textual inversion on SDXL is complicated. The base images were generated from SDXL itself.

Notes

It's possible to use not wrong in the prompt itself but in testing it has no effect.
You can use other negative prompts in conjunction with the wrong prompt but you may want to weight them.