Cognenza
EN

AI-assisted translation from the Spanish original. Read the original

Tools · Level intermediate

ComfyUI: what it is and how to start creating AI images on your computer

A beginner’s guide to ComfyUI and Comfy Desktop: official requirements, installation on Windows and Mac, nodes and workflows, offline use, privacy and limits.

By Cognenza editorial team · Updated · 4 min read · Sources cited

What it is

ComfyUI is an open-source program for creating images, video, audio and 3D models with AI on your own computer. Instead of a text box, it uses a canvas of nodes: each box does one thing (load a model, write the description, generate, save) and you join them with wires to form a workflow. That way you can repeat and adjust exactly how each image is made.

Comfy Desktop is the app that installs and manages ComfyUI for you: Python, dependencies and updates included.

Who it’s for

  • Yes: people who want full control, to work offline or to use open-weight models without paying per image.
  • Better something else: if you just want a quick image, the assistants in free AI images are simpler.

Official requirements

Comfy Desktop
Windows10 or later, x64 or ARM64. Dedicated NVIDIA/AMD GPU recommended, not required
macOS13 Ventura or later, Apple Silicon (M1 or later)
LinuxOfficial .deb and AppImage packages
Memory8 GB of RAM minimum, 16 GB recommended
DiskAt least 4.85 GB per installation, plus the models you download

Step-by-step installation

  1. Download Comfy Desktop from comfy.org/download.
  2. Windows: run the .exe installer and open it from the Start menu. Mac: open the .dmg and drag the app to Applications.
  3. On the first screen choose Local (installed on your computer, works offline) or Cloud (paid cloud GPUs).
  4. Check the boxes: accepting the terms is required; sharing anonymous usage data and beta features are ticked and you can untick them.
  5. Open a text-to-image template, write your description and press Ctrl+Enter. Save the workflow with Ctrl+S.

Precautions

  • Custom nodes: they extend ComfyUI, but they are third-party code running on your computer. Install only reputable ones and check where they come from.
  • Models: download only from reliable sources and respect their licence and terms of use.
  • Real people: don’t generate images of someone without their permission. The EU AI Act requires disclosing when content is a deepfake.

Limits

  • There is a learning curve: complex workflows take practice.
  • Speed depends heavily on your graphics card and memory.
  • Models take up a lot of disk space.

Frequently asked questions

Is ComfyUI free?

Yes. ComfyUI is open source (GPL-3.0 licence) and runs on your computer. Comfy Cloud and the API nodes for commercial models are paid and optional.

Do I need a graphics card?

A dedicated GPU (NVIDIA or AMD) is recommended on Windows for good performance, but it is not required. On Mac it works with Apple Silicon (M1 or later).

Does it work without internet?

Yes. According to its repository, the core doesn’t download anything unless you ask it to, and the --offline option disables the paid API nodes so everything stays on your computer.

Does it send usage data?

During first setup, Comfy Desktop asks whether you want to share anonymous usage data; the box is ticked and you can untick it.

Keep reading

Sources

Official sources are linked in their original language.

  1. Comfy Desktop: visión general — Comfy Org
  2. Requisitos del sistema — Comfy Org
  3. Comfy Desktop para Windows — Comfy Org
  4. Repositorio oficial Comfy-Org/ComfyUI — GitHub

Change log

  • : First published with the official documentation and version 0.40.0 (13 October 2026).

Educational content, prepared with the help of AI tools, with sources cited. It is not a substitute for professional advice. Tools change fast: if you spot something out of date, write to us at contacto@cognenza.com. How we work.