Running LM Studio and ComfyUI on one GPU

The post explains how the author runs LM Studio, ComfyUI, and OpenWebUI on one AI server with a single GPU. The setup uses a VRAM cleanup node in ComfyUI, LM Studio server settings, and a GPU memory offload setting. The author says it works on a Geforce 5060ti 16GB system with 64GB of memory and Bazzite Linux.

Key points

  • Install LM Studio and ComfyUI on the same system.
  • Configure OpenWebUI so it can talk to both tools.
  • Add the VRAM cleanup node to the ComfyUI workflow.
  • Turn on LM Studio’s setting that limits model offload to dedicated GPU memory.
  • If it fails, check that the workflow unloads the model before image output.

Quick term guide

LM Studio
A simple application that lets you download and run large AI models on your own computer.
OpenWebUI
A web interface for using AI models and tools in one place.
local AI agent
An AI program that runs on your own computer instead of a remote cloud server
local AI
AI software that runs entirely on your own computer, with no internet connection needed.
AI agents
AI agents are AI tools that can carry out steps toward a goal, not just answer once.
AI agent
An AI program that can inspect information and suggest what to do next.
image generation
Image generation means creating pictures from written instructions.
workflow
A repeatable set of steps for getting a task done.
Read original