Wan Workbench
Category

ComfyUI Under the Hood

Node-by-node workflow breakdowns, dependency setup, graph design, custom nodes, model loading, and practical debugging for ComfyUI.

2 articles

Page 1 of 1 | Total 2 articles

How to Read a Wan2.2 ComfyUI Workflow Without Guessing
ComfyUI Under the Hood |

How to Read a Wan2.2 ComfyUI Workflow Without Guessing

Learn how to systematically read and break down a Wan2.2 ComfyUI workflow instead of guessing. This guide maps out the five core stages—model loading, prompt encoding, sampling, VAE decode, and video assembly—helping you isolate variables, trace settings, and achieve fully reproducible AI video generation results on consumer hardware.

2
Which Node Is Actually Using All Your VRAM?
ComfyUI Under the Hood |

Which Node Is Actually Using All Your VRAM?

This article explains how to identify which node actually consumes your VRAM during AI video generation on consumer hardware like the RTX 3090. By replacing assumptions with a systematic isolation method—testing loaders, VAE decodes, and frame counts one variable at a time—it helps creators turn memory anxiety into predictable, reproducible workflows.

3