Sdxs 512 Dreamshaper

🚀 SDXS-512-DreamShaper

SDXS is a model capable of generating high-resolution images in real-time based on prompt texts. It is trained using score distillation and feature matching. For detailed information, please refer to our research paper: SDXS: Real-Time One-Step Latent Diffusion Models with Image Conditions. We have open-sourced this model as part of our research.

SDXS-512-DreamShaper is a version specifically trained for the community. This model is trained without emphasizing FID, sacrificing diversity for better image generation quality. To avoid potential risks, SDXS-512-1.0 and SDXS-1024-1.0 will not be available soon. Stay tuned to our repo for updates.

✨ Features

• Generate high-resolution images in real-time based on prompt texts.
• Trained using score distillation and feature matching.
• Specifically trained for the community, sacrificing diversity for better image quality.

📦 Installation

The installation steps are not explicitly provided in the original document. However, the usage example implies that you need to have the diffusers library installed. You can install it using the following command:

pip install diffusers

💻 Usage Examples

Basic Usage

import torch
from diffusers import StableDiffusionPipeline, AutoencoderKL

repo = "IDKiro/sdxs-512-dreamshaper"
seed = 42
weight_type = torch.float16     # or float32

# Load model.
pipe = StableDiffusionPipeline.from_pretrained(repo, torch_dtype=weight_type)
pipe.to("cuda")

prompt = "a close-up picture of an old man standing in the rain"

# Ensure using 1 inference step and CFG set to 0.
image = pipe(
    prompt, 
    num_inference_steps=1, 
    guidance_scale=0,
    generator=torch.Generator(device="cuda").manual_seed(seed)
).images[0]

image.save("output.png")

📚 Documentation

Model Information

Compatibility Issue

Similar to SDXS-512-0.9, since our image decoder is not compatible with diffusers, we use TAESD. Currently, our pull request has been merged to reduce the gap between TAESD and our image decoder. In the next diffusers release update, we may replace the image decoder.

📄 License

The model is licensed under the OpenRAIL++ license.

📚 Cite Our Work

@article{song2024sdxs,
  author    = {Yuda Song, Zehao Sun, Xuanwu Yin},
  title     = {SDXS: Real-Time One-Step Latent Diffusion Models with Image Conditions},
  journal   = {arxiv},
  year      = {2024},
}