The original model, weights, and documentation are developed and maintained by Stability AI.
The model weights hosted here are unmodified and redistributed as-is.
This model is released under the CreativeML Open RAIL++-M License. Users are responsible for complying with all applicable license terms and use-based restrictions.
This repository is not affiliated with, sponsored by, or endorsed by Stability AI.
Model
SDXL consists of an ensemble of experts pipeline for latent diffusion:
In a first step, the base model is used to generate (noisy) latents,
which are then further processed with a refinement model (available here: https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/) specialized for the final denoising steps.
Note that the base model can be used as a standalone module.
Alternatively, we can use a two-stage pipeline as follows:
First, the base model is used to generate latents of the desired output size.
In the second step, we use a specialized high-resolution model and apply a technique called SDEdit (https://arxiv.org/abs/2108.01073, also known as "img2img")
to the latents generated in the first step, using the same prompt. This technique is slightly slower than the first one, as it requires more function evaluations.
Model Description: This is a model that can be used to generate and modify images based on text prompts. It is a Latent Diffusion Model that uses two fixed, pretrained text encoders (OpenCLIP-ViT/G and CLIP-ViT/L).
For research purposes, we recommend our generative-models Github repository (https://github.com/Stability-AI/generative-models), which implements the most popular diffusion frameworks (both training and inference) and for which new functionalities like distillation will be added over time.
Clipdrop provides free SDXL inference.
1from diffusers import DiffusionPipeline
2import torch
3pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, use_safetensors=True, variant="fp16")4pipe.to("cuda")5# if using torch < 2.06# pipe.enable_xformers_memory_efficient_attention()7prompt ="An astronaut riding a green horse"8images = pipe(prompt=prompt).images[0]
To use the whole base + refiner pipeline as an ensemble of experts you can run:
py
1from diffusers import DiffusionPipeline
2import torch
3# load both base & refiner4base = DiffusionPipeline.from_pretrained(5"stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, variant="fp16", use_safetensors=True6)7base.to("cuda")8refiner = DiffusionPipeline.from_pretrained(9"stabilityai/stable-diffusion-xl-refiner-1.0",10 text_encoder_2=base.text_encoder_2,11 vae=base.vae,12 torch_dtype=torch.float16,13 use_safetensors=True,14 variant="fp16",15)16refiner.to("cuda")17# Define how many steps and what % of steps to be run on each experts (80/20) here18n_steps =4019high_noise_frac =0.820prompt ="A majestic lion jumping from a big stone at night"21# run both experts22image = base(23 prompt=prompt,24 num_inference_steps=n_steps,25 denoising_end=high_noise_frac,26 output_type="latent",27).images
28image = refiner(29 prompt=prompt,30 num_inference_steps=n_steps,31 denoising_start=high_noise_frac,32 image=image,33).images[0]
When using torch >= 2.0, you can improve the inference speed by 20-30% with torch.compile. Simple wrap the unet with torch compile before running the pipeline:
For more information on how to use Stable Diffusion XL with diffusers, please have a look at the Stable Diffusion XL Docs.
Uses
Direct Use
The model is intended for research purposes only. Possible research areas and tasks include
Generation of artworks and use in design and other artistic processes.
Applications in educational or creative tools.
Research on generative models.
Safe deployment of models which have the potential to generate harmful content.
Probing and understanding the limitations and biases of generative models.
Excluded uses are described below.
Out-of-Scope Use
The model was not trained to be factual or true representations of people or events, and therefore using the model to generate such content is out-of-scope for the abilities of this model.
Limitations and Bias
Limitations
The model does not achieve perfect photorealism
The model cannot render legible text
The model struggles with more difficult tasks which involve compositionality, such as rendering an image corresponding to “A red cube on top of a blue sphere”
Faces and people in general may not be generated properly.
The autoencoding part of the model is lossy.
Bias
While the capabilities of image generation models are impressive, they can also reinforce or exacerbate social biases.
