How to beginner · 3 min read

How to install Stable Diffusion

Quick answer

To install Stable Diffusion, set up a Python environment with Python 3.8+, then install the diffusers and transformers libraries via pip. Use the StableDiffusionPipeline from diffusers to load the model and generate images.

PREREQUISITES

Python 3.8+
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117 (for CUDA GPU)
pip install diffusers transformers scipy ftfy accelerate
Hugging Face account with access token (for model download)

Setup

Install Python 3.8 or higher. For GPU acceleration, install PyTorch with CUDA support. Then install the diffusers library and its dependencies. You also need a Hugging Face account to access the Stable Diffusion model weights.

bash

python3 -m venv sd-env
source sd-env/bin/activate
pip install --upgrade pip
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117
pip install diffusers transformers scipy ftfy accelerate
huggingface-cli login

Step by step

Use the following Python script to load the Stable Diffusion pipeline and generate an image from a prompt. Replace YOUR_HF_TOKEN with your Hugging Face access token if needed.

python

import os
from diffusers import StableDiffusionPipeline
import torch

# Set your Hugging Face token as environment variable
os.environ["HUGGINGFACE_TOKEN"] = os.environ.get("HUGGINGFACE_TOKEN")

model_id = "runwayml/stable-diffusion-v1-5"

pipe = StableDiffusionPipeline.from_pretrained(
    model_id,
    use_auth_token=os.environ["HUGGINGFACE_TOKEN"],
    torch_dtype=torch.float16
).to("cuda")

prompt = "A fantasy landscape, vivid colors"
image = pipe(prompt).images[0]

image.save("output.png")
print("Image saved as output.png")

output

Image saved as output.png

Common variations

For CPU-only machines, remove torch_dtype=torch.float16 and .to("cuda").
Use pipe.enable_attention_slicing() to reduce VRAM usage on GPUs with limited memory.
Try other Stable Diffusion versions like stabilityai/stable-diffusion-2 by changing model_id.
For asynchronous usage, wrap calls with asyncio and use pipe(prompt) inside async functions.

Troubleshooting

If you get authentication errors, ensure your Hugging Face token is valid and set in HUGGINGFACE_TOKEN environment variable.
On CUDA errors, verify your GPU drivers and CUDA toolkit versions match the installed PyTorch version.
If out of memory errors occur, enable attention slicing or reduce image resolution.
For slow performance on CPU, consider using a smaller model or running on GPU.

✅

Key Takeaways

Use Python 3.8+ and install PyTorch with CUDA for best performance.
Install the diffusers library and authenticate with Hugging Face to access Stable Diffusion models.
Run the StableDiffusionPipeline to generate images with simple Python code.
Enable attention slicing to reduce GPU memory usage if needed.
Check environment variables and GPU setup to troubleshoot common errors.

Verified 2026-04 · runwayml/stable-diffusion-v1-5, stabilityai/stable-diffusion-2

Verify ↗