ComfyUI: understand, install, troubleshoot

What ComfyUI is, how to install it on a production workstation, what nodes, extensions and models are, and how to diagnose a failure, seen from the Axiris studio, with the official documentation as the reference.

ComfyUI is a piece of software that represents an image or video generation as a graph of nodes: each operation, whether loading a model, encoding the prompt, sampling, decoding, applying a LoRA, correcting an area or upscaling, is visible, connected to the next and replayable identically. That is what makes it a production environment rather than one more application. At Axiris, ComfyUI is the environment where workflows are built and measured; this section explains the software itself, from installation to troubleshooting, drawing on the official documentation.

This section is for those who want to run ComfyUI themselves: the technical creative, the graphic designer setting up a workstation, the professional who has inherited a machine to understand. It explains the software. Knowing when to go through it in your work is another question, covered in When to go through ComfyUI in a workflow.

What ComfyUI is

ComfyUI represents a generation process as a graph of nodes. Each node is an operation: load a model, encode the prompt, produce the latent, sample, decode, apply a LoRA, guide with ControlNet, transform an image, correct an area, upscale, process a video. Connected to each other, these operations form a workflow you can see in full.

Where a classic interface hides the process behind a text field and a button, ComfyUI shows it. For a creative, this changes one essential thing: they can understand what is happening, change one step without touching the others, and replay the same process identically.

A workflow can be simple:

prompt → model → generation → image

or chain several steps:

references → generation → transformation → mask → local correction → upscaling → finishing

The point is not to use the largest number of nodes. The point is to build the workflow that answers the project.

What Axiris does with it

At Axiris, ComfyUI is not one generation tool among others. It is the environment in which workflows are built, measured and frozen: we no longer generate an image, we build it, step by step, and what is validated no longer moves. The graph is what makes this promise tenable, because it replays the same operations in the same order.

For a creative, this takes two forms: their workflow, built with them on their own projects, and a production workstation they understand and run themselves. For a business, a third: a machine in Switzerland, rented by the hour and delivered ready, with ComfyUI and Nova installed. Nothing to buy, nothing to administer.

The official documentation and the studio's experience

The official ComfyUI documentation remains the reference for how the software works, its versions and its installation procedures; the source code is public: Comfy-Org / ComfyUI. Every page in this section refers to it.

Axiris adds what the documentation does not say: choosing a method according to the project, integrating it into a graphic designer's work, configuring a production workstation, using recent models, diagnosing real failures.

What we have published

What the studio publishes on ComfyUI is dated and tested on its own machines: DLSS 5 in ComfyUI: we publish our pack, and what we have understood (10 October 2026) and Comfy Agent: AI becomes the natural interface of the workflow (6 October 2026). For the place of ComfyUI in a graphic designer's work: the guide Integrating AI into a graphic designer's workflow without losing control.

A question about ComfyUI?

The Synnexia ComfyUI AI answers questions on installation, configuration, nodes and troubleshooting; its corpus includes the official documentation. Give it an error, a console excerpt or a workflow to explain.

To build your workflow or set up your workstation: Contact request: a first conversation, free and with no commitment. Every practice is different; it is by listening to you that we propose what suits you.