---
title: Ep 3: Nodes 101
slug: ep-3-nodes-101
createdAt: 2026-06-12T05:56:16.046Z
---



::embed[]{url="https://www.youtube.com/watch?v=119Ui0pUFXQ&list=PLQeGICOxlVdeoe-HqquUEWLtGHWCI0ZA7&index=4"}

## Adding nodes

Two ways to add a node:

- **Right-click** the canvas to browse by category
- **Double-click** to search by name (faster once you know what you're looking for)

## Common nodes

**Load Image** brings an image into your workflow. Drag and drop a file or browse your computer.

**CLIP Text Encode** is where you type your prompt. It converts your text into a format the AI model can understand.

**Load Diffusion Model** selects which AI model your workflow uses and loads it into memory.

**Reroute** redirects connection lines to keep your canvas tidy. It doesn't do anything to the data.

## The color system

Every connector on a node is color-coded. The color tells you what type of data it carries. You can only connect matching colors.

| Color                                       | Type         | What it is                              |
| ------------------------------------------- | ------------ | --------------------------------------- |
| <font color="#14B8A6">**Green**</font>****  | String       | Text. Prompts, file paths, descriptions |
| <font color="#0F766E">**Blue**</font>       | Image        | Visible images (RGB pixels)             |
| <font color="#6B21A8">**Violet**</font>     | Model        | Diffusion model files                   |
| <font color="#FBCFE8">**Pink**</font>       | Latent       | Data in the AI's internal workspace     |
| <font color="#EF4444">**Red**</font>        | VAE          | Encoder/decoder model                   |
| <font color="#FEF08A">**Yellow**</font>     | CLIP         | Text translation model                  |
| <font color="#F97316">**Orange**</font>**** | Conditioning | Processed prompt data                   |
| <font color="#A5F3FC">**Turquoise**</font>  | ControlNet   | Structural control model                |



:::hint{type="success"}
Connectors on the left side of a node are inputs. Right side are outputs. Data always flows left to right.
:::



## Reading a workflow

Once you know the colors and the direction, you can read any workflow. It doesn't matter how complex it looks. Follow the lines from left to right, check the colors, and you can trace exactly how data moves through the system.

***

## FAQ



### What is latent space?

It's where the AI does its work internally. Latent data looks like noise to us. The AI constructs images in this space, then a VAE decode node converts the result into a visible image.

### What if the colors don't match?

You can't connect them. ComfyUI prevents it. If you need to pass data between incompatible types, you need a conversion node in between (like VAE Encode to go from blue/image to pink/latent).

### What is conditioning?

It's your prompt after it's been processed by CLIP. The raw text goes in, conditioning comes out, and that's what actually guides the AI during generation.

### Why is my node blank?

A blank node usually means it hasn't received data yet. Check that all required inputs on the left side are connected. If a connector is empty, the node has nothing to work with and will show a blank state.


::::VerticalSplit{layout="middle"}
:::VerticalSplitItem
[Ep 2: Run your first workflow](docId\:SqWpz7z7vreMU34iAHXrJ)        &#x20;
:::

:::VerticalSplitItem
&#x20;            [Ep 4: Text-to-Image from scratch](docId\:p7Cv7o3yZIW6DVnGg5vEv)
:::
::::