With the increasing popularity of AI painting and video generation tools like Stable Diffusion and ComfyUI, more and more creators are looking for their own dedicated AI creation environment. Compared to local computers, which are limited by graphics card performance and storage space, deploying AI tools on cloud-based Windows GPU servers is becoming a more flexible and cost-effective option.

This article will demonstrate in its entirety how to deploy Stable Diffusion WebUI and ComfyUI on a Windows GPU cloud server, covering everything from environment preparation to model loading.

Why choose a cloud-based Windows GPU server?

The biggest hurdle to deploying AI painting locally is hardware. An RTX 4090 graphics card can cost tens of thousands of yuan, and models with 16GB or more of VRAM are even more expensive. Cloud GPU servers offer a more flexible solution:

Pay-as-you-go, shut down when not in use. Local graphics cards are non-refundable once purchased, while cloud instances are billed hourly, making costs controllable during intermittent creation.

Flexible configuration, upgrade anytime. From T4 to A100, from 16GB to 80GB of VRAM, you can switch at any time according to project needs, without incurring hardware depreciation.

Easy to get started with the Windows system. No need to learn Linux command line; the remote desktop connection provides an experience almost identical to using a local computer.

Jtti GPU cloud servers are designed for this type of scenario. Equipped with NVIDIA data center-grade GPUs, high-frequency CPUs, and NVMe SSDs, the Hong Kong node uses a bidirectional CN2 GIA direct connection to mainland China, resulting in extremely low latency for remote desktop operations.

Environment Preparation: Three things to do after receiving the server

Step 1: Confirm GPU and Driver

After logging into the Windows GPU cloud server via remote desktop, open Task Manager to check if the GPU is correctly recognized. Run `nvidia-smi` in PowerShell to confirm the graphics card model, memory size, and driver version. Jtti's GPU instances come pre-installed with NVIDIA GRID drivers, ready to use out of the box.

Step 2: Install Python Environment

AI drawing tools rely on a Python runtime environment. Python version 3.10.6 is recommended—it's the most stable version with the most compatibility with the Stable Diffusion WebUI. Be sure to check "Add Python to PATH" during installation.

Step 3: Install Git

Both Stable Diffusion WebUI and ComfyUI pull code from Git. Download the installation package from the Git website and complete the installation with the default options.

Once the environment is ready, you can begin deployment.

Option 1: Stable Diffusion WebUI (Qiuye Integration Package) – A Beginner's Choice

Stable Diffusion WebUI is currently the most popular AI drawing tool, with an intuitive browser interface and a rich plugin ecosystem.

The Qiuye Integration Package bundles the Python environment, dependency libraries, and WebUI together, requiring no command-line operations, making it especially suitable for beginners.

Step 1: Install Startup Dependencies

Double-click to run “Startup Dependencies-dotnet-6.0.11.exe” and wait for the installation to complete.

Step 2: Unzip the Integration Package

Unzip sd-webui-aki-v4.11.1-cu128.zip to the target directory (D drive or a non-system drive is recommended).

Step 3: Place Model Files

- Place the downloaded large model file (.ckpt or .safetensors format) into the `\sd-webui-aki-v4\models\Stable-diffusion` directory.

- Place the ControlNet model into the `\sd-webui-aki-v4\models\ControlNet` directory.

Step 4: Launch WebUI

Double-click “Launcher.exe”, and click “One-Click Launch” in the interface. The first launch will automatically download dependencies; please wait patiently. Your browser will automatically open `http://127.0.0.1:7860`.

System Requirements: It is recommended to have an NVIDIA GPU with ≥8GB of VRAM and 100-200GB of reserved hard drive space for model files. Jtti's GPU cloud servers come standard with NVMe SSDs, providing significantly faster model loading speeds than ordinary hard drives.

Option 2: ComfyUI – A Workflow Tool for Advanced Users

ComfyUI organizes the drawing process using node connections, precisely controlling each step, making it suitable for advanced users requiring complex workflows. Method 1: Comfy Desktop (Simplest)

The official Comfy Desktop version is available, offering a command-line-free installation with built-in Python and PyTorch environments.

Download Comfy-Desktop-Setup.exe and double-click to install. It's recommended to choose a non-system drive (e.g., D drive). Initial launch requires an internet connection to download basic dependencies; subsequent launches allow for completely offline image generation. After opening, select "Local" mode and follow the wizard to complete the configuration.

Method 2: Git Cloning Installation (Suitable for Customization Needs)

The original ComfyUI is installed via Git and the Python command line. In PowerShell, execute:

git clone https://github.com/comfyanonymous/ComfyUI.git

cd ComfyUI

Create a Python virtual environment in the ComfyUI directory and install dependencies. After completion, run `main.py` to start the service. Access `http://127.0.0.1:8188` in your browser.

Method 3: One-Click Installation Script (Recommended for Technical Users)

A one-click installation script for ComfyUI on Windows is available on GitHub, automatically handling the installation of CUDA, PyTorch, and custom nodes. After downloading the repository, double-click `install_update_comfyui.bat` to complete the installation. Starting the script `launch_comfyui.bat` will automatically detect crashes and restart.

Core Advantages of Cloud Deployment: Dual Guarantee of Computing Power and Storage

Deploying AI tools in the cloud offers two easily overlooked advantages besides the flexibility of pay-as-you-go pricing:

Ample Storage Space. Local hard drives are often filled with model files, while cloud servers can elastically expand storage. Jtti's GPU cloud servers come standard with NVMe SSDs, providing extremely high efficiency in model loading and data read/write.

Remote Access and Team Collaboration. ComfyUI or WebUI deployed in the cloud can be accessed externally via a public IP address and port. Team members do not need to deploy individually; they can collaborate using a shared environment.

The barrier to entry for deploying AI painting and video generation tools on Windows is rapidly decreasing. The AI ​​integration package allows beginners to complete their first image in just 10 minutes, while the ComfyUI desktop version lets advanced users say goodbye to the command line. Cloud GPU servers completely eliminate hardware limitations—you don't need to spend tens of thousands of yuan on a graphics card; you can have professional-grade AI creation capabilities simply by renting an hourly account.

Jtti cloud servers in Hong Kong offer bidirectional CN2 GIA direct connections to mainland China, with remote desktop latency as low as 12-18ms. Whether it's Stable Diffusion WebUI or ComfyUI, the out-of-the-box Windows environment allows you to focus on the creative process itself, rather than struggling with environment configuration.