How to Run Stable Diffusion on a VPS: An Affordable Guide for AI Enthu

Affiliate disclosure: We earn commissions when you shop through the links on this page, at no additional cost to you.

Sam Torres
AI Automation & Self-Hosting Specialist

Stable Diffusion has revolutionized the world of AI art generation, allowing anyone to create stunning images from simple text prompts. While online services offer convenience, many developers and AI enthusiasts seek greater control, privacy, and cost-effectiveness by running Stable Diffusion on their own Virtual Private Server (VPS).

This guide will walk you through the entire process of setting up Stable Diffusion on an affordable VPS, with a special focus on Contabo, a provider known for its high-performance yet budget-friendly servers. We’ll cover everything from choosing the right VPS specifications to installation and optimization, ensuring you can unleash your creativity without breaking the bank.

Why Run Stable Diffusion on a VPS?

Running Stable Diffusion locally on your personal machine can be resource-intensive, often requiring a powerful GPU that many do not possess. Cloud-based services offer a solution, but they can quickly become expensive, especially with frequent or extended use. A VPS strikes a perfect balance, providing dedicated resources and flexibility at a fraction of the cost of premium cloud GPU instances.

Cost-Effectiveness: Avoid the upfront cost of expensive hardware or the recurring fees of high-tier cloud platforms. Contabo, in particular, offers powerful hardware at competitive prices, making it an excellent choice for long-term projects.
Control & Privacy: You have full root access to your server, allowing for complete customization of your environment. Your data and generated images remain private, away from third-party servers.
Dedicated Resources: Unlike shared hosting, a VPS guarantees a certain amount of CPU, RAM, and storage, ensuring consistent performance for your Stable Diffusion tasks.
Accessibility: Access your Stable Diffusion instance from anywhere, any device, through a web interface. You can set up projects and check progress remotely.
Scalability: As your needs grow, most VPS providers, including Contabo, allow you to easily upgrade your server specifications.

Choosing the Right VPS for Stable Diffusion

The success and speed of your Stable Diffusion setup heavily depend on your VPS specifications. Here’s what to look for:

GPU vs. CPU Inference

Stable Diffusion thrives on GPU power. If your budget allows, a VPS with a dedicated GPU (even an entry-level one) will provide significantly faster image generation times compared to CPU-only inference. However, GPU-equipped VPS instances are generally more expensive. If you’re on a very tight budget or only generate images occasionally, a powerful multi-core CPU can still handle Stable Diffusion, albeit slower.

Contabo offers excellent VPS plans, including options with significant CPU resources that can be surprisingly effective for CPU-based inference. For those serious about AI, their dedicated servers or higher-tier Cloud VPS instances can be configured with GPUs, providing the necessary horsepower.

RAM and Storage

RAM: Stable Diffusion models can consume a fair amount of RAM. For optimal performance, especially when loading larger models or generating high-resolution images, aim for at least 16GB of RAM. 32GB or more is ideal.
Storage: The Stable Diffusion models themselves can be several gigabytes, and you’ll accumulate many generated images. Opt for fast NVMe SSD storage, with at least 100GB to start. More is always better if you plan to keep a large archive of models and creations.

Operating System

Ubuntu LTS (Long Term Support) is the recommended operating system due to its widespread community support, ease of package management, and excellent compatibility with AI frameworks like PyTorch and CUDA (for GPU acceleration).

Contabo: The Go-To for Affordable AI VPS

When it comes to balancing performance and price, Contabo consistently stands out. They offer a range of VPS and dedicated server solutions that are particularly well-suited for resource-intensive tasks like running Stable Diffusion.

Why Contabo?

Unbeatable Value: Contabo provides robust hardware configurations at prices that often undercut competitors. This is crucial for AI workloads that demand substantial computing power.
High-Performance Hardware: Their servers feature fast AMD EPYC or Intel Xeon processors and NVMe SSDs, ensuring quick model loading and inference.
Generous Resources: Even their mid-range VPS plans come with ample CPU cores and RAM, making CPU-only Stable Diffusion feasible and efficient.
Scalability: Easily upgrade your VPS plan as your Stable Diffusion projects grow in complexity and demand more resources.

Recommended Contabo Plans for Stable Diffusion

For most users looking to get started, a Contabo Cloud VPS with sufficient RAM is a great starting point. If you plan to heavily utilize Stable Diffusion or other AI models, consider their dedicated servers for maximum performance and potential GPU integration.

Contabo Plan	Typical Specs (approx.)	Stable Diffusion Suitability
Cloud VPS S	4 vCPU, 8GB RAM, 50GB NVMe	Entry-level for CPU-only inference, slow. Good for testing/learning.
Cloud VPS M	6 vCPU, 16GB RAM, 100GB NVMe	Recommended minimum for CPU-only inference. Decent performance.
Cloud VPS L	8 vCPU, 30GB RAM, 200GB NVMe	Excellent for CPU-only inference, faster image generation.
Cloud VPS 60	10 vCPU, 60GB RAM, 1.6TB NVMe	Premium CPU-only performance. If GPUs are not available, this is your best bet for general AI workloads. Check out the Contabo Cloud VPS 60 here.
Dedicated Servers	High-end CPU, significant RAM, optional GPUs	Optimal for heavy AI workloads and GPU acceleration. Explore Contabo Dedicated Servers.

You can find more details on their offerings and choose a plan that suits your needs on the Contabo homepage.

Setting Up Your Contabo VPS for Stable Diffusion

Once you’ve selected and provisioned your Contabo VPS, it’s time for the technical setup. We’ll assume you’ve chosen an Ubuntu 22.04 LTS image.

Step 1: Initial Server Setup

Connect to your VPS via SSH:

ssh root@YOUR_VPS_IP_ADDRESS

Update your system:

sudo apt update && sudo apt upgrade -y

Install essential tools:

sudo apt install -y build-essential python3-dev python3-venv git curl

Step 2: Install NVIDIA Drivers and CUDA (GPU VPS ONLY)

If you have a GPU-equipped VPS, this step is critical. Skip if you are performing CPU-only inference.

First, identify your GPU and recommended drivers:

ubuntu-drivers devices

Install the recommended driver (e.g., nvidia-driver-535):

sudo apt install -y nvidia-driver-535

Reboot for drivers to take effect:

sudo reboot

After rebooting, SSH back in and verify the installation:

nvidia-smi

This should display your GPU information. Next, install CUDA Toolkit. Find the correct version compatible with your Stable Diffusion setup. For most cases, a recent CUDA version from NVIDIA’s official site is best. Here’s an example for CUDA 11.8 (adjust as needed):

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/11.8.0/local_installers/cuda-repo-ubuntu2204-11-8-local_11.8.0-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2204-11-8-local_11.8.0-1_amd64.deb
sudo cp /var/cuda-repo-ubuntu2204-11-8-local/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cuda

Add CUDA to your PATH (add these lines to your ~/.bashrc or ~/.profile):

export PATH="/usr/local/cuda-11.8/bin:${PATH}"
export LD_LIBRARY_PATH="/usr/local/cuda-11.8/lib64:${LD_LIBRARY_PATH}"

Reload your shell:

source ~/.bashrc

Step 3: Install Stable Diffusion (e.g., Automatic1111 Web UI)

We’ll use the popular Automatic1111 web UI for Stable Diffusion due to its ease of use and extensive features.

Clone the repository:

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

Create a Python virtual environment:

python3 -m venv venv
source venv/bin/activate

Run the web UI. The first run will download all necessary dependencies and the default Stable Diffusion model:

python3 launch.py --listen --share

The --listen flag makes the web UI accessible from your VPS’s IP address. The --share flag generates a Gradio link, which is useful for temporary sharing or if you have strict firewall rules. For persistent access, it’s better to configure a reverse proxy (e.g., Nginx) and open port 7860 on your firewall.

Alternatively, you can run it without --share and access it directly via http://YOUR_VPS_IP_ADDRESS:7860.

Allow traffic on port 7860 (Ubuntu UFW firewall example):

sudo ufw allow 7860/tcp
sudo ufw enable

Once the setup is complete, you should see a URL in your terminal where the Stable Diffusion web UI is running. Open this URL in your browser, and you’re ready to start generating images!

Optimizing Performance

To get the most out of your Stable Diffusion setup on Contabo, consider these optimizations:

Model Pruning/Quantization: Use smaller or optimized versions of Stable Diffusion models (e.g., fp16 versions) which consume less VRAM (on GPU) or RAM (on CPU) and can speed up inference.
Leverage Command-Line Arguments: The launch.py script accepts various arguments to optimize performance, such as --xformers (for NVIDIA GPUs), --medvram, or --lowvram to reduce memory usage. For CPU inference, explore arguments like --use-cpu-backend if available or simply ensure your Python/PyTorch installation is optimized for CPU.
Regular Updates: Keep your Stable Diffusion web UI, Python libraries, and operating system updated to benefit from performance improvements and bug fixes.
Background Processes: Ensure no unnecessary background processes are consuming valuable CPU or RAM resources on your VPS.
Dedicated User: Run Stable Diffusion under a non-root user for better security and resource management.

Troubleshooting Common Issues

“CUDA out of memory” error: If you’re using a GPU, this means your VRAM is insufficient. Try using --medvram or --lowvram flags in launch.py, or opt for smaller models.
Slow generation: Check your VPS resources. For CPU inference, more cores and RAM will help. For GPU, ensure drivers and CUDA are correctly installed and detected by nvidia-smi.
Web UI not accessible: Verify that port 7860 is open in your VPS firewall (sudo ufw status) and that --listen is used in launch.py. Check if the process is actually running (ps aux | grep stable-diffusion).
Dependencies not installing: Ensure you are within your Python virtual environment (source venv/bin/activate) and have build-essential installed.

Conclusion

Running Stable Diffusion on an affordable VPS like those offered by Contabo is a fantastic way to gain control over your AI art generation workflow. You get dedicated resources, flexibility, and privacy without the hefty price tag of premium cloud services or the upfront investment in high-end local hardware.

By carefully selecting your VPS, configuring your environment, and applying the optimization tips outlined in this guide, you can establish a powerful and cost-effective platform for all your AI creative endeavors. Dive in and start generating your masterpieces!

Disclaimer: This article contains affiliate links to Contabo. We may earn a commission if you make a purchase through these links, at no extra cost to you. This helps support AIStackDigest.

What to Read Next

Bookmark aistackdigest.com for daily AI tools, reviews, and workflow guides.

This article was produced with the assistance of AI tools and reviewed by the AIStackDigest editorial team.

How to Run Stable Diffusion on a VPS: An Affordable Guide for AI Enthusiasts