ComfyUI turns a text prompt into an image through connected steps
Priya N. Chen on 14th Nov 2026
ComfyUI turns a text prompt into an image through connected steps
ComfyUI is an open-source, node-based tool that lets artists build a workflow for making an image from text. A node is a block that performs one job. The maker’s beginner tutorial uses Stable Diffusion 1.5, an image-making model, to produce a 512 × 512 image from a text description. A model is a system trained on examples to produce results from inputs. The workflow has seven nodes across six types, added to a blank canvas and connected like steps in a recipe.
The process starts with Load Checkpoint, which loads the model file. Two CLIP Text Encode (Prompt) nodes turn text into information the model can use. A prompt is text that guides the result. One node takes what the artist wants to see; the other takes what the artist wants to leave out. Empty Latent Image sets the image size and makes an early form of image data. KSampler uses the model and prompts to shape that data. VAE Decode turns it into a visible image, and Save Image stores the result. The artist supplies the text, connects the stages and decides what to keep.
To follow the tutorial on a computer, a beginner installs ComfyUI through Comfy Desktop, ComfyUI Portable or a manual setup. The tutorial says to download the Stable Diffusion model file and place it in a folder ComfyUI can read. Then the artist adds and connects the nodes, enters prompts, sets the image size, runs the workflow and saves the image. A Comfy Cloud template offers a browser-based starting point without local installation. ComfyUI describes its self-hosted version as free. Cloud has a free tier and paid subscription plans, but the documentation gives no prices.
Is AI a tool in the hand of the artist, as the camera and the brush were, or something else? The hope is faster image exploration and more ways to shape a picture. The fear is lost commissions, training without consent or payment, work flattened into style, and power gathering in a few firms. ComfyUI’s documentation does not state what training data was used or give details about consent or credit. Meredith Whittaker’s recorded stance is surveillance and concentrated industry power. I read that concern as relevant to artists weighing a cloud-based workflow. The model handles image-making steps, but the artist’s choices still guide the prompt, workflow and final selection.
Join Neoclassical Pop Art daily take on the arts, the weekly letter, and first word on product launches, sales, and free gifts.