ComfyUI on a GPU server for image and video generation
Build image and video generation workflows from nodes, with the checkpoints, LoRAs and custom nodes you choose, and keep queues running while you sleep. Run it on a GPU server with the interface behind HTTPS and a login, preinstalled as an app option or set up with our guide; GPU servers are being added to our range, so ask us about availability.
- Node-based workflows for images and video
- Your checkpoints, LoRAs and custom nodes
- GPU servers being added: ask about availability
- Interface behind HTTPS and a login
- Stack
- Python 3.13, PyTorch
- Default portComfyUI has no authentication by default: expose it only through a reverse proxy that requires a login.
- 8188 (127.0.0.1 by default)
- MinimumComfyUI’s README says its memory management runs large models on as little as 4 GB of VRAM and 8 GB of RAM; more VRAM is faster.
- From 4 GB VRAM and 8 GB RAM
- Your data
- models/, output/, custom_nodes/
- Official docs
- docs.comfy.org
Workflows with
- Checkpoints
- LoRA
- ControlNet
- Custom nodes
- ComfyUI-Manager
- API workflows
- Multi-user
- Open WebUI
Facts from the project’s official website, documentation and repository, checked in October 2026.
GPU servers are being added
GPU servers are being added to our range step by step. Tell us the models or workloads you plan to run, how many users you expect and your preferred location, and we will let you know what is available and when.
Which server size fits?
Starting points for vCPU, memory and disk. Grow the server when your data and users grow.
| Feature |
Learning
Trying workflows, slow on CPU
|
Recommended Image generation
One GPU for a creator or a small team
|
Video and large models
More GPU memory, long queues
|
|---|---|---|---|
| GPU | None (--cpu) | One GPU | One or more GPUs |
| MemoryMemory for the app, its database and the operating system. | 8–16 GB | 16–32 GB | 64 GB+ |
| vCPUVirtual processor cores of the server. | 4 | 8 | 16 |
| DiskCheckpoints take several GB each; outputs grow quickly. | 50 GB | 200 GB | 500 GB+ |
| Server type | VDS or dedicated | GPU server (ask us) | GPU server (ask us) |
-
Learning
Trying workflows, slow on CPU
- GPU
- None (--cpu)
- MemoryMemory for the app, its database and the operating system.
- 8–16 GB
- vCPUVirtual processor cores of the server.
- 4
- DiskCheckpoints take several GB each; outputs grow quickly.
- 50 GB
- Server type
- VDS or dedicated
-
Recommended
Image generation
One GPU for a creator or a small team
- GPU
- One GPU
- MemoryMemory for the app, its database and the operating system.
- 16–32 GB
- vCPUVirtual processor cores of the server.
- 8
- DiskCheckpoints take several GB each; outputs grow quickly.
- 200 GB
- Server type
- GPU server (ask us)
-
Video and large models
More GPU memory, long queues
- GPU
- One or more GPUs
- MemoryMemory for the app, its database and the operating system.
- 64 GB+
- vCPUVirtual processor cores of the server.
- 16
- DiskCheckpoints take several GB each; outputs grow quickly.
- 500 GB+
- Server type
- GPU server (ask us)
ComfyUI’s README gives 4 GB of VRAM and 8 GB of RAM as enough to run even large models with its memory management; the system requirements page gives no fixed figures. CPU mode works but is very slow.
Your image pipeline, always on
Design once in the graph, then run it from the browser or the API.
Node-based workflows
Combine models, LoRAs, ControlNet and upscalers in a graph, and share workflows as files.
ComfyUI-Manager
Install, update and manage custom nodes from the interface when you start ComfyUI with --enable-manager.
API and integrations
Queue workflows over the API from your app, or connect Open WebUI to generate images in chat.
Multi-user storage
Start ComfyUI with --multi-user to give each user their own storage.
Your data stays yours
Prompts, files and databases stay on a server you control, in the location you choose, instead of a shared SaaS account.
Full root access
Install what the app needs, change any setting and run more services next to it. Nothing is locked behind a panel.
Help from our team
Open a ticket from the client area if the server or the network needs attention, and follow every reply in one place.
From order to first login
Order the app preinstalled on your server, or install it yourself with our guide.
-
Ask about a GPU server
Tell us the models and resolutions you work with; we reply with what is available and its price.
-
Add ComfyUI
Select ComfyUI as an app option when you order, or install it on a clean Ubuntu or Debian server with our guide.
-
Add models and nodes
Copy checkpoints into models/checkpoints and install the custom nodes your workflows need.
-
Protect the interface
Keep ComfyUI on 127.0.0.1 and publish it through a reverse proxy with HTTPS and a login.
Step-by-step setup guides
Install, secure and update the app with our guides, written for current Ubuntu and Debian releases.
-
How to install ComfyUI on an NVIDIA GPU server securely
Set up ComfyUI from the official repository on a GPU server, run it under its own user with systemd, keep it off the public internet, add HTTPS with basic authentication, and manage models, custom nodes, backups and updates.
45 min Intermediate
Related solutions
Frequently asked questions
Does ComfyUI need a GPU?
For real work, yes. ComfyUI runs on NVIDIA GPUs with CUDA, AMD GPUs with ROCm, Intel Arc and Apple silicon, and has a --cpu mode that is much slower. Its README says it can run large models on as little as 4 GB of VRAM and 8 GB of RAM, but more GPU memory means faster and larger images.
Is ComfyUI safe to open to the internet?
Not as it is. The ComfyUI server has no authentication by default, and custom nodes run code on the server. Keep it on 127.0.0.1 and put a reverse proxy with HTTPS and a login in front of it, or reach it over a VPN.
Is there an official Docker image?
No. ComfyUI’s documentation states that it does not provide an official Docker image, so we install it in a Python environment with the PyTorch build for your GPU, which is the documented manual installation.
How do I manage custom nodes?
Start ComfyUI with --enable-manager to use ComfyUI-Manager for installing and updating custom nodes. Its security levels block risky installs when the server listens on a public address; install only nodes you trust.
How much disk space do I need?
Checkpoints are several gigabytes each, video models more, and outputs pile up quickly. Plan at least 200 GB of NVMe storage for regular image work and clean up the output folder on a schedule.
Can I use ComfyUI from my app?
Yes. Export a workflow in API format and queue it over ComfyUI’s HTTP and WebSocket API from your code, or let Open WebUI use ComfyUI for image generation in chat.
Ask about a GPU server for ComfyUI
Tell us your models, resolutions and how many images or videos you generate, and we will tell you what fits and what is available.