Course: ComfyUI — Generating Images and Video with Node-Based Workflows

Topics covered

• •Choose the right installation method (Desktop, Portable, manual, cloud) for your hardware and OS.
• •Understand the system requirements and ComfyUI's folder structure (especially `models/`).
• •Learn the main interface elements: the canvas, sidebar, menu, template library, and keyboard shortcuts.

Topics covered

• •Load a ready-made text2img workflow three ways: from templates, from a PNG's metadata, and from JSON.
• •Install a checkpoint model (SD1.5/SDXL) both automatically and manually.
• •Run your first generation and understand the role of every node in the base graph.

Topics covered

• •Understand the concept of a workflow as a node graph and visual programming.
• •Learn how a node is built: inputs, outputs, widgets, properties, and modes (mute/bypass).
• •Master links, the color coding of data types, and techniques for organizing a graph.

Topics covered

• •Master img2img and the role of the denoise parameter.
• •Learn inpaint/outpaint using the mask editor.
• •Connect one or several LoRAs for stylization.

Topics covered

• •Understand how a video workflow differs from an image graph (a latent with a time axis, frame count, FPS).
• •Get to know the main open video models: Wan 2.x, Hunyuan Video, LTX-Video.
• •Master text2video and image2video modes, plus control modes (first-last frame, control).

Topics covered

• •Understand the difference between interpolated and neural network upscaling (ESRGAN and similar).
• •Master the basic Upscale Image (using Model) pipeline and the text2img + upscale combo.
• •Get to know latent upscaling and the hires fix technique (a second sampler pass).

Topics covered

• •Learn to install custom nodes (Manager, git, Registry) and manage dependencies.
• •Understand how a custom node is built: a Python class, INPUT_TYPES, a function, a category.
• •Understand how the ComfyUI server works: HTTP routes, WebSocket, the queue.