On August 10, 2026, a ComfyUI tutorial was published detailing how to run FLUX, a default open-weight image model of 2026, locally in 13 steps. The guide explains that running FLUX locally removes reliance on external APIs, rate limits, and content filters, with generation costing nothing beyond electricity. The tutorial covers confirming GPU VRAM requirements, installing Python, Git, and GPU drivers, cloning and installing ComfyUI, launching it, installing ComfyUI Manager, selecting the right FLUX model for hardware, downloading matching VAE and text encoder files, organizing model folders, loading a workflow template, configuring prompts and sampler settings, generating the first image, adding a LoRA for style or character consistency, and saving workflows for automation via the ComfyUI API. The process takes about 90 minutes total, mostly spent on model downloads. The result is a working local FLUX pipeline with a configured ComfyUI install, a downloaded model matched to the GPU, a saved workflow, and a script for queuing new generations without a browser. The tutorial notes that a diffusion pipeline requires a VAE and text encoders alongside the model, and that no machine learning background is needed, only patience with file paths, a GPU with enough headroom, and willingness.