Install Stable Diffusion WebUI locally on Ubuntu Linux

It is possible to run ATOMATIC1111’s Stable Diffusion WebUI locally on Linux PC or deploy it to servers.

1. Environment

OS：Ubuntu 22.04 LTS
Kernel：5.19.0
CPU: Intel i5-7400
GPU: Nvidia GTX1050 Ti 4GB
RAM: 16GB
20GB+ disk space
Stable Diffusion WebUI version：0cc0ee1 (2023/2/20)

2. Download Stable Diffusion models

Go to HuggingFace or Civitai to find a model.

Use wget command to download the model

1
wget https://huggingface.co/andite/anything-v4.0/blob/main/anything-v4.5-pruned.ckpt

Here are some good models for your reference.

2.1. Realistic style models

Stable Diffusion, the original model published by CompVis and StabilityAI.

Chilloutmix

Deliberate

2.2. Anime style models

I would suggest you to start from “Anything” model if you want to draw anime artworks.

3. Install Stable Diffusion WebUI

3.1. Install dependencies

Install proprietary Nvidia drivers in order to use CUDA. Then reboot.

1
2
3
4
5
sudo apt update
sudo apt purge *nvidia*
# List available drivers for your GPU
ubuntu-drivers list
sudo apt install nvidia-driver-525

Follow the instructions on Nvidia Developers to install CUDA. Reboot again.

1
2
3
4
5
6
7
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-ubuntu2204.pin
sudo mv cuda-ubuntu2204.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/12.1.0/local_installers/cuda-repo-ubuntu2204-12-1-local_12.1.0-530.30.02-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2204-12-1-local_12.1.0-530.30.02-1_amd64.deb
sudo cp /var/cuda-repo-ubuntu2204-12-1-local/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cuda

Verify the installation

1
2
nvidia-smi
nvcc --version

Install Python, wget, git

1
sudo apt install python3 python3-pip python3-virtualenv wget git

Because we need a Pyhon 3.6 enviroment for SD WebUi, we have to install Anaconda

1
2
3
4
wgt https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
chmod +x Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh
# Confirm everything and install it

Create a virtual environment of Python 3.10.6

1
conda create --name sdwebui python=3.10.6

3.2. Clone Stable Diffusion WebUI repository

Clone the repository of Stable Diffusion WebUI

1
2
cd ~
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

Move .ckpt models to stable-diffusion-webui

1
mv ~/anything-v4.5-pruned.ckpt  ~/stable-diffusion-webui/models/Stable-diffusion/

Enter the virtual enviroment

1
conda activate sdwebui

If you want to activate virtual environment in a bash script, add these on the top of webui-user.sh

1
2
eval "$(conda shell.bash hook)"
conda activate sdwebui

3.3. Setup commandline arguments

According to Wiki，we have to change some commdanline arguments in order to start SD WebUI.

Edit webui-user.sh

1
vim webui-user.sh

If the VRAM of GPU is lower than 4GB, add: COMMANDLINE_ARGS=--medvram --opt-split-attention

If your PC has RAM lower than 8GB, add: COMMANDLINE_ARGS=--lowvram --opt-split-attention

You could also add --listen so you can access the WebUI from other PC on the same network. Or add --share to generate a public Gradio link for accessing WebUI while deploying SD WebUI to servers.

3.4. Launch Stable Diffusion WebUI

Run webui.sh, it will install all the dependencies. Then a link should pop up: http://127.0.0.1:7860

1
./webui.sh

To access WebUI from other PC on the same network，enter http://<IP of the PC>:7860 in the address bar of your browser. Don’t forget to open firewall port

1
2
sudo ufw allow 7860/tcp
sudo ufw reload

3.5. How to update Stable Diffusion WebUI

Get current branch

1
git rev-parse --short HEAD

Pull latest files

1
git pull

If something is broken after updating, roll back to the previous branch

1
git checkout "branch"

4. How to use Stable Diffusion WebUI

4.1. Prompts

Use “Prompts” and “Ngative Prompts” to tell AI what to draw.

See Vodly Artist name and Danbooru tags for choosing prompts.

For example, to draw Jeanne from Fate/Grand Order, we type the name of the character and characteristics of her body in the prompt fields.

1
jeanne d'arc from fate grand order, girl, (best quality), (masterpiece), (high detail), ((full face)), sharp, ((looking at viewer)), ((detailed pupils)), (thick thighs), (((full body))), (large breasts), vagina, nude, nipples

Then type negative prompts.

1
lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts,signature, watermark, username, blurry, artist name

4.2. Text to image

Go to SD WebUI, type the prompts
Check Restore faces
Clcik Generate butoon, it will start generate a image
You shall see the result at the right panel

All generated images will be stored at stable-diffusion-webui/outputs

You can also increase the value of Batch count so it will generate multiple images in one run.

4.3. Imge to image

Type the prompts
Upload a image. Check Restore faces. Clcik Generate.

You can change the value of CFG Scale and Denoising strength. The lower the value of Denoising strength is, the output would be more similar the original image.

Click Interrogate Deepboooru to generate prompts automatically accodring to the image you uploaded.

Install Stable Diffusion WebUI locally on Ubuntu Linux

Related Posts: