Deploy Automatic1111 Stable Diffusion on RamNode VPS

Prerequisites

Recommended VPS Specs

Automatic1111 is GPU-accelerated but can run in CPU-only mode. Here is what to plan for:

Use Case	RAM	vCPUs	Storage	Notes
CPU-only (testing)	8 GB	4	40 GB	Slow — 60–300s per image
CPU-only (workable)	16 GB	8	80 GB	Better for batch jobs
GPU-accelerated	16 GB+	4+	80 GB+	Requires GPU VPS

Storage note: Base Stable Diffusion v1.5 checkpoint is ~4 GB. SDXL models are 6–7 GB each. Budget at least 40 GB for a single-model setup, more if you plan to collect LoRAs, VAEs, and multiple checkpoints.

What You Need Before Starting

A RamNode KVM VPS running Ubuntu 22.04 LTS
Root or sudo SSH access
A domain or your VPS IP for remote access
Basic familiarity with the Linux command line

Update System & Install Dependencies

Update system packages

apt update && apt upgrade -y

Install required system packages

apt install -y \
  wget curl git python3 python3-pip python3-venv \
  libgl1 libglib2.0-0 libsm6 libxext6 libxrender-dev \
  build-essential libssl-dev libffi-dev python3-dev \
  ffmpeg libopencv-dev bc

Python 3.10+ ships with Ubuntu 22.04, which satisfies Automatic1111's requirements. Verify:

Verify Python version

python3 --version

Create a Dedicated User

Running Automatic1111 as root is not recommended. Create a dedicated user:

Create sduser

adduser --disabled-password --gecos "" sduser
usermod -aG sudo sduser
su - sduser

All remaining steps run as sduser unless noted otherwise.

Install Automatic1111

Clone the repository

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

GPU Acceleration (NVIDIA, if applicable)

If your RamNode VPS has an NVIDIA GPU, install CUDA Toolkit 11.8:

Install CUDA Toolkit

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.0-1_all.deb
sudo dpkg -i cuda-keyring_1.0-1_all.deb
sudo apt update
sudo apt install -y cuda-toolkit-11-8

Verify GPU detection

nvidia-smi

CPU-Only Mode

No additional setup is needed for CPU-only. The launch script handles the rest with the --skip-torch-cuda-test flag covered in Step 6.

Download a Stable Diffusion Model

Automatic1111 requires at least one checkpoint model to start. Place models in ~/stable-diffusion-webui/models/Stable-diffusion/

Option A — Stable Diffusion v1.5 (~4 GB)

Download SD v1.5

cd ~/stable-diffusion-webui/models/Stable-diffusion/
wget -O v1-5-pruned-emaonly.safetensors \
  "https://huggingface.co/runwayml/stable-diffusion-v1-5/resolve/main/v1-5-pruned-emaonly.safetensors"

Option B — SDXL Base 1.0 (~6.5 GB, higher quality)

Download SDXL Base

cd ~/stable-diffusion-webui/models/Stable-diffusion/
wget -O sd_xl_base_1.0.safetensors \
  "https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors"

Tip: Use safetensors format over .ckpt whenever available — it loads faster and has no arbitrary code execution risk.

Configure the Launch Script

Create a webui-user.sh file to customize startup flags:

Prepare launch script

cd ~/stable-diffusion-webui
cp webui-user.sh.example webui-user.sh 2>/dev/null || touch webui-user.sh
nano webui-user.sh

GPU Setup (NVIDIA)

webui-user.sh (GPU)

#!/bin/bash
export COMMANDLINE_ARGS="--listen --port 7860 --enable-insecure-extension-access"

CPU-Only Setup

webui-user.sh (CPU-only)

#!/bin/bash
export COMMANDLINE_ARGS="--listen --port 7860 --skip-torch-cuda-test --precision full --no-half --enable-insecure-extension-access"

Make executable

chmod +x webui-user.sh

Note: --listen binds to 0.0.0.0 so the UI is accessible remotely. You will secure this with SSH tunneling or a reverse proxy in Step 8.

First Launch

Run the web UI for the first time. This will take 5–15 minutes as it sets up a Python virtual environment and downloads dependencies (~5 GB):

Launch the web UI

cd ~/stable-diffusion-webui
./webui.sh

Watch the output for the line:

Expected output

Running on local URL:  http://0.0.0.0:7860

Once you see this, the web UI is running. Leave this terminal session open (or use screen/tmux to background it).

Secure Access

Port 7860 should not be exposed publicly without authentication. Use one of these methods:

Option A — SSH Tunnel (Simplest)

From your local machine:

Create SSH tunnel

ssh -L 7860:localhost:7860 sduser@YOUR_VPS_IP

Then open http://localhost:7860 in your browser. The UI is only accessible through your SSH session.

Option B — Nginx Reverse Proxy with Password Auth

Install Nginx and create password

sudo apt install -y nginx apache2-utils
sudo htpasswd -c /etc/nginx/.htpasswd yourusername

/etc/nginx/sites-available/automatic1111

server {
    listen 80;
    server_name YOUR_DOMAIN_OR_IP;

    auth_basic "Restricted";
    auth_basic_user_file /etc/nginx/.htpasswd;

    location / {
        proxy_pass http://127.0.0.1:7860;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "upgrade";
        proxy_set_header Host $host;
        proxy_read_timeout 300;
    }
}

Enable and reload

sudo ln -s /etc/nginx/sites-available/automatic1111 /etc/nginx/sites-enabled/
sudo nginx -t && sudo systemctl reload nginx

Add HTTPS with Let's Encrypt

Obtain TLS certificate

sudo apt install -y certbot python3-certbot-nginx
sudo certbot --nginx -d yourdomain.com

Run as a Persistent Service

Keep Automatic1111 running after SSH disconnect using systemd:

/etc/systemd/system/automatic1111.service

[Unit]
Description=Automatic1111 Stable Diffusion Web UI
After=network.target

[Service]
Type=simple
User=sduser
WorkingDirectory=/home/sduser/stable-diffusion-webui
ExecStart=/home/sduser/stable-diffusion-webui/webui.sh
Restart=on-failure
RestartSec=10

[Install]
WantedBy=multi-user.target

Enable and start the service

sudo systemctl daemon-reload
sudo systemctl enable automatic1111
sudo systemctl start automatic1111

Check status

sudo systemctl status automatic1111
journalctl -u automatic1111 -f

Managing Models and Storage

Check disk usage

du -sh ~/stable-diffusion-webui/models/*
df -h

Model Directory Structure

Directory layout

models/
├── Stable-diffusion/   # Main checkpoint files (.safetensors, .ckpt)
├── Lora/               # LoRA weights
├── VAE/                # VAE files
├── ControlNet/         # ControlNet models
└── embeddings/         # Textual inversion embeddings

Reload Models Without Restarting

In the web UI, go to Settings → Actions → Reload UI or click the refresh button next to the model dropdown to pick up new files dropped into the models directory.

Performance Tuning

Reduce VRAM/RAM Usage

Add these flags to COMMANDLINE_ARGS in webui-user.sh for lower-memory systems:

GPU memory optimization flags

--medvram          # Use medium VRAM mode (GPU setups)
--opt-split-attention  # Reduce peak VRAM usage
--no-half-vae      # Fix black image issues with some VAEs

For CPU-only on a RAM-constrained server:

CPU memory optimization

--lowvram --always-batch-cond-uncond

Xformers (GPU only, significant speed boost)

Install xformers

sudo -u sduser pip install xformers

Add --xformers to COMMANDLINE_ARGS.

Troubleshooting

Issue	Solution
Black/corrupted images	Add `--no-half-vae` to launch args
CUDA out of memory	Add `--medvram` or `--lowvram` to launch args
Very slow on CPU	Expected — reduce resolution to 512×512 and lower steps to 20
Port 7860 unreachable	Verify `--listen` is set and check `sudo ufw status`
webui.sh exits immediately	Check `journalctl -u automatic1111 --since "5 minutes ago"`

Automatic1111 Deployed Successfully!

Your self-hosted Stable Diffusion Web UI is now running on a RamNode VPS. Explore these extensions to expand functionality:

ControlNet — Precise image composition control
ADetailer — Automatic face/hand touch-up on generated images
Civitai Helper — Browse and download models from Civitai directly in the UI
sd-webui-regional-prompter — Multi-subject compositions with region-specific prompts

Install extensions via Extensions → Install from URL inside the web UI.