COMMUNITY PAGE
ComfyUI for Beginners
ComfyUI 101: A Beginner's Guide
ComfyUI is a node-based interface for building and running AI workflows. This guide introduces the essential concepts - from understanding nodes and connections to generating images, upscaling, and using ControlNet - so you can start building workflows with confidence.
ComfyUI 101: A Beginner's Guide
ComfyUI is a node-based interface for building and running AI workflows. This guide introduces the essential concepts - from understanding nodes and connections to generating images, upscaling, and using ControlNet - so you can start building workflows with confidence.
ComfyUI 101: A Beginner's Guide
ComfyUI is a node-based interface for building and running AI workflows. This guide introduces the essential concepts - from understanding nodes and connections to generating images, upscaling, and using ControlNet - so you can start building workflows with confidence.
What this guide covers:
1. What Is ComfyUI?
Get familiar with ComfyUI, understand its interface, and see how node-based workflows make AI generation possible.
ComfyUI is a visual tool for building AI workflows using nodes. Each node performs a particular task, and connections between nodes determine how information moves through the workflow.
A workflow is a collection of connected nodes that work together to produce an output. For example, a text-to-image workflow uses a model and a prompt to generate an image:
Load Model → Process Prompt → Generate Image → Decode → Output
This is a simplified illustration. A real workflow may contain additional nodes and connections depending on the model being used.
Why Use ComfyUI?
ComfyUI allows you to:
- Customize how an AI workflow operates.
- Modify individual components and generation settings.
- Work with text and image inputs.
- Reuse and organize workflow components.
- Explore techniques such as upscaling and ControlNet.
You don't need to understand every node before getting started. Begin with a simple workflow and learn how its components connect.
2. Run Your First Workflow
Ready to try it yourself? Follow along to run a workflow and see how the different steps come together.
The easiest way to understand ComfyUI is to start with an existing workflow. A template provides a ready-made graph with the required nodes and connections. You can supply a prompt, adjust supported settings, and run the workflow without building everything yourself.
Steps to Run a Workflow
- Open a text-to-image workflow.
- Locate the prompt input and enter your description.
- Review the available settings.
- Click Run to start generation.
- Wait for the workflow to finish and inspect the output.
For example, you could enter:
A small cabin in a snowy forest at sunrise,
soft golden light, cinematic photography
The resulting image depends on the model, prompt, and workflow settings. Running a template first helps you understand the process before you start creating your own graphs.
3. Understand Nodes and Connections
See how nodes communicate with each other and how connecting them creates a complete workflow.
Nodes are the building blocks of a ComfyUI workflow. Each node has inputs and outputs that determine how it interacts with other nodes.
- Inputs receive information from other components.
- Outputs pass information to connected components.
- Connections show how information flows through the workflow.
For example, a model loader provides a model to another node, while a prompt-conditioning node provides information derived from the text prompt.
Understanding Connector Types
ComfyUI uses color-coded connectors to distinguish different data types. Inputs and outputs must be compatible for a connection to work. If a connection cannot be created, check whether the connector types match.
When exploring an existing workflow, follow the connections from the input nodes through the processing stages to the final output.
Every lesson in this guide maps to a hands-on workflow in the ComfyUI 101 course on Floyo. Open the workflow, follow along, and run it in your browser - no local install needed.
4. Build a Text-to-Image Workflow
Turn a text prompt into an image and explore the key steps behind the generation process.
A text-to-image workflow generates an image from a written description. Although the exact graph depends on the model, a typical workflow includes several stages:
| Stage | Purpose |
|---|---|
| Model loading | Loads the model used for generation |
| Prompt conditioning | Converts the prompt into information the model can use |
| Latent preparation | Provides the latent input required by the workflow |
| Sampling | Generates a latent result using the model and sampling settings |
| Decoding | Converts the latent representation into an image |
| Output | Displays or saves the generated image |
This diagram describes the general process rather than an importable workflow. Different models may require different nodes, loaders, or processing steps.
When building a workflow, focus on understanding what each node receives, what it does, and what it sends to the next component.
5. Understand KSampler Settings
Explore the settings that influence your results and learn how small adjustments can change the generated image.
KSampler is used in many diffusion-based workflows to control the sampling process. Its settings affect how generation is performed and can influence the final output.
| Setting | Description |
|---|---|
| Seed | Controls the starting noise used during generation. Reusing a seed can help reproduce results when other conditions remain the same. |
| Steps | Sets the number of sampling steps. More steps can increase processing time and do not always improve the result. |
| CFG | Controls how strongly generation follows the prompt conditioning. Higher values are not always better. |
| Sampler | Selects the sampling algorithm used by the workflow. |
| Denoise | Controls the amount of denoising applied. Its effect depends on the workflow and input. |
Tips for Beginners
- Start with the settings provided by a working workflow.
- Change one parameter at a time.
- Compare outputs to understand the effect of each change.
- Remember that appropriate settings depend on the model and workflow.
On Floyo, every lesson workflow is pre-built and ready to run. Change KSampler settings, swap models, and compare outputs without installing anything on your machine.
6. Organize Workflows With Subgraphs
Discover how to group related nodes, simplify complex workflows, and keep your workspace organized.
As workflows become more complex, the canvas can become difficult to navigate. Subgraphs allow you to group related nodes into a single block. This helps organize the canvas without losing the underlying workflow operations.
Subgraphs are useful for:
- Grouping nodes that perform a related task.
- Reducing visual clutter.
- Making large workflows easier to navigate.
- Reusing grouped components where supported.
It's best to understand individual nodes first, then use subgraphs to organize larger workflows.
7. Understand Models and Supporting Components
Learn what different model components do and how they work together to produce your final output.
ComfyUI workflows use models and supporting components to generate and process images. Understanding their roles makes it easier to read workflows and identify what each component contributes.
| Component | Description |
|---|---|
| Diffusion model | Performs the core image-generation process in a diffusion-based workflow. |
| LoRA | Adapts a compatible base model toward a particular style, subject, or concept. |
| CLIP | Helps encode text prompts into representations used by supported workflows. |
| VAE | Encodes or decodes between image space and latent space, depending on the workflow. |
| ControlNet | Uses additional structural information to guide generation. |
Not every workflow uses all these components in the same way. The model architecture and workflow determine which components are required.
8. Work With Image-to-Image Workflows
Start with an existing image and see how prompts and workflow settings can help transform it.
Image-to-image workflows use an existing image as an input. Instead of starting only from a text prompt, you can use a photograph, sketch, or other visual reference as the starting point for a transformation.
| Text-to-Image | Image-to-Image |
|---|---|
| Starts primarily from a text prompt | Starts with an image and may also use a prompt |
| Generates an image from the prompt and workflow | Transforms or modifies an existing visual |
| Useful for creating new images from descriptions | Useful when working from a visual reference |
The amount of change depends on the model, workflow, and settings. Some workflows preserve more of the original image, while others allow more substantial transformations.
Floyo has pre-loaded models, LoRAs, and image-to-image templates you can open and run directly. Upload your reference image, adjust settings, and generate - no model downloads or folder management.
9. Understand Image Upscaling
Explore how upscaling workflows increase image resolution and help enhance visual details.
Upscaling increases an image's resolution. Depending on the method, it may also reconstruct or enhance visual details. Upscaling can be useful when an image needs to be larger for a design project, display, or other output requirement.
Things to Consider
- Output resolution: A larger image contains more pixels but does not guarantee more accurate details.
- Upscaling method: Different approaches may enlarge an image or use AI to reconstruct details.
- Input quality: The original image affects the final result.
- Processing requirements: Higher-resolution workflows may require more processing time and memory.
Check the final output for artifacts or unwanted details, especially when using AI-based enhancement.
10. Guide Generation With ControlNet
See how ControlNet uses structural guidance, such as poses, edges, or depth, to influence image generation.
ControlNet allows supported workflows to use structural information to guide image generation. This can provide more control over composition than relying on a text prompt alone. The course introduces four techniques:
| Technique | Input | Purpose |
|---|---|---|
| Canny | Edge information | Helps guide outlines and structural boundaries |
| Depth | Depth information | Helps guide spatial relationships |
| OpenPose | Pose information | Helps guide body and limb positioning |
| Line Art | Line drawings | Helps guide the structure of the generated image |
For example, if you want to guide the pose of a generated character, an OpenPose-based workflow can use pose information as additional guidance.
The result depends on the input reference, model, and workflow settings. ControlNet helps guide generation but does not guarantee an exact match.
ControlNet workflows need specific models and preprocessor nodes. On Floyo, Canny, Depth, OpenPose, and Line Art workflows are pre-configured with the required models loaded - open one and start experimenting.
Learn ComfyUI on Floyo
Learning ComfyUI doesn't have to start with a local installation. On Floyo, the ComfyUI 101 course workflows run in your browser, so you can focus on understanding the concepts and following the lessons.
This is useful if you're new to ComfyUI and want to learn the fundamentals before managing local installations, dependencies, and model files.
You can follow the course in order, practise with the examples, and apply the same concepts when working with ComfyUI elsewhere.
Frequently asked questions
Is ComfyUI suitable for beginners?
Yes. Although ComfyUI has a learning curve, beginners can start with a template and gradually learn about nodes, connections, and workflow settings.
Do I need coding experience?
No. The basic workflow-building process is visual. Some advanced workflows may require additional technical knowledge, but coding isn't necessary to learn the fundamentals.
Do I need my own GPU?
You don't need your own GPU to follow the course workflows on Floyo, which run in your browser. Running ComfyUI locally has separate hardware requirements.
Is ComfyUI the same as Stable Diffusion?
No. Stable Diffusion is a family of AI models, while ComfyUI is a tool for running models and building workflows. ComfyUI supports multiple models.
Where should I start?
Begin with the first lesson to understand ComfyUI, then run a template and learn how nodes connect. Continue through the course to explore workflow building, models, image-to-image, upscaling, and ControlNet.
Conclusion
Now that you understand the basic concepts behind ComfyUI, you can continue through the lessons and practise each technique with the course workflows.