Cloud Running Guide

Run ComfyUI on cloud services for GPU-free usage.

Cloud Platforms

Recommended Platforms

Platform	GPU Options	Price	Difficulty
Vast.ai	RTX 3090, A100	$0.2-0.5/hr	Medium
RunPod	RTX 3090, A100	$0.3-0.6/hr	Medium
Modal	Various	Pay-as-you-go	Medium
Google Colab	T4 (Free)	Free/Limited	Easy

Vast.ai

Setup Steps

1. Create account at https://vast.ai
2. Add credits (minimum $10)
3. Search for "ComfyUI" templates
4. Rent GPU and start

Recommended Config

Config	Value
GPU	RTX 3090 or better
CUDA	12.1+
Disk	100GB+
Price	< $0.4/hr

Connect via SSH

# Connect to instance
ssh root@<instance-ip>

# Clone ComfyUI
git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI

# Install dependencies
pip install -r requirements.txt

# Run
python main.py --listen 0.0.0.0

RunPod

Using Templates

1. Create account at https://runpod.io
2. Search "ComfyUI" in templates
3. Deploy with one click
4. Access via provided URL

Google Colab

Free Option

# Install ComfyUI
!git clone https://github.com/comfyanonymous/ComfyUI.git
%cd ComfyUI
!pip install -r requirements.txt

# Run
!python main.py --listen

Access

Click the localtunnel URL to access ComfyUI interface.

Model Upload

Upload to Cloud

# Using scp (Vast.ai/Rental)
scp model.safetensors root@<ip>:/root/ComfyUI/models/checkpoints/

# Using rclone (Google Drive)
rclone copy model.safetensors remote:comfyui-models/

Download from URLs

import urllib.request

url = "https://model-url.com/model.safetensors"
filename = url.split("/")[-1]
urllib.request.urlretrieve(url, f"/path/to/{filename}")

Cost Optimization

Stop When Not Using

Action	Savings
Stop instance after use	100% of idle cost
Use smaller GPU for testing	50% cost
Use spot instances	70% cost

Efficient Workflow

1. Test on smaller images first
2. Use lower steps for preview
3. Final output on better GPU

Security

Important Notes

• Never expose SSH keys publicly
• Use strong passwords
• Enable firewall when possible
• Delete instances after use

Data Safety

• Back up important workflows locally
• Don't store sensitive data on cloud
• Delete instances to avoid charges

Troubleshooting

Q: Connection timeout

A: Check if ComfyUI is running and port is correct. Ensure firewall allows the port.

Q: Out of memory

A: Use smaller batch size or lower resolution. Choose GPU with more VRAM.

Q: Instance disconnected

A: Save work frequently. Use cloud storage for important files.

Next Steps

• Installation Guide - Local installation
• Model Installation - Install models