The following uses the https://github.com/ostris/ai-toolkit GitHub repository to train a local LoRA on user-provided images. We will then create images using the generated LoRA with ComfyUI. Running this tool requires an Nvidia GPU with 24GB of VRAM. We will train on Ubuntu 24.04 with a recent Nvidia driver installed, git, brew (to install useful commands), and Python (python3 with pip3 and the venv package installed, either via apt or brew)
![FLUX.1[dev] with ComfyUI and Stability Matrix (Rev: 08/24)](
https://imagedelivery.net/IEMzXmjRvW0g933AN5ejrA/wwwnotionso-image-prod-files-secures3us-west-2amazonawscom-ed175516-6d97-436e-aa46-cdf22facffbd-aa5c4a72-2145-4e6e-977c-ecabdeee3fc4-20240810-flux1devjpg/public
)
In this post, we will run FLUX.1 [dev] within ComfyUI, proposing two installation methods: 1) with Dockge to run the tool within a docker compose; a server-compatible installation method, 2)using Stability Matrix, a desktop application with easy model integration with multiple inference engines available.
This post details the build as a container of Automatic1111 and its integration as an image generator option for the “Ollama with Open WebUI” installation.
