Offload image upscaling to a remote server with a dedicated GPU. Useful when your main ComfyUI machine has limited VRAM or when you want to use a different GPU for upscaling tasks.

Features

Remote upscaling via HTTP API
Real-time progress bar in ComfyUI
Support for .pth, .pt, and .safetensors upscale models
Configurable tile size for memory management
Model caching on server for faster subsequent runs

Installation

ComfyUI Node (Client)

Copy the remote_upscale folder to your ComfyUI custom_nodes directory
Restart ComfyUI
The "Remote Upscale" node will appear in the "image/upscaling" category

Upscale Server

Copy the server folder to your remote machine
Install dependencies:
```
pip install -r requirements.txt
```
Install PyTorch with CUDA/ROCm support appropriate for your GPU

Set the models path and run:

export UPSCALE_MODELS_PATH="/path/to/your/upscale_models"
python server.py

The server will start on port 8189 by default.

Usage

Start the upscale server on your remote machine
In ComfyUI, add the "Remote Upscale" node
Set the server_url to your server address (e.g., http://192.168.1.100:8189)
Select a model from the dropdown
Connect an image input and run the workflow

Node Parameters

server_url: URL of the remote upscale server
model_name: Upscale model to use (fetched from server)
tile_size: Tile size for processing large images (default: 512)
show_progress: Enable progress bar during upscaling

Refresh Models

Use the "Refresh Remote Models" node to update the model list from the server.

Server API

Endpoint	Method	Description
`/models`	GET	List available models
`/upscale`	POST	Upscale image (no progress)
`/upscale_stream`	POST	Upscale with SSE progress
`/result/{id}`	GET	Fetch result after streaming
`/health`	GET	Server health check

Requirements

Client (ComfyUI)

requests

Server

fastapi
uvicorn
python-multipart
torch (with CUDA or ROCm)
torchvision
spandrel
pillow
numpy

License

MIT