ComfyUI Guide: Managing Workflows, Models, and Configuration

ComfyUI is a free, locally-run application that lets you generate AI images and videos on your own hardware. It is designed for users who desire greater control than a standard prompt-based interface offers, allowing you to visualize and fine-tune every stage of the generation pipeline. While this flexibility introduces a learning curve, this guide is structured to support your journey from initial setup through to executing and customizing your first workflow.

Prerequisites for Getting Started

To operate ComfyUI effectively, your computer must possess sufficient GPU resources to handle the specific models and workflows you plan to execute. Higher-complexity models and more intensive workflows typically demand greater VRAM capacity.

Additionally, you will need access to the specific model files required by your workflow. Depending on the model architecture, these files may include checkpoints, diffusion models, VAEs, text encoders, LoRAs, or other components. These assets are generally located within the ComfyUI/models directory.

If your local hardware lacks the necessary GPU capabilities, you have the option to run ComfyUI on a remote GPU-enabled desktop. This approach offloads the computational burden of generation to the remote GPU, allowing you to manage the process from your regular workstation.

Setting Up ComfyUI

For users on Windows and macOS, the ComfyUI team recommends the desktop application as the optimal entry point. While other installation paths exist, such as manual setup or using the ComfyUI command-line tool, the best approach depends on your specific operating system and environment.

Once installation is complete, launch the application to access the interface. Here, you will encounter the workflow canvas along with the essential tools for creating and managing your projects.

The Importance of ComfyUI Workflows

A ComfyUI workflow dictates the precise methodology for generating images or videos. It orchestrates the selection of models, the configuration of settings, and the sequence of processing steps required to achieve the final output.

This structure grants you significantly more agency than a simple text prompt box. You have the ability to swap models, integrate LoRAs, utilize input images, modify generation parameters, upscale outputs, or insert additional processing stages.

Furthermore, workflows are designed for preservation and reuse. Rather than recreating your configuration from scratch each time, you can store a workflow that yields satisfactory results and adjust specific parameters as needed. You can also leverage workflows shared by the community and adapt them to fit your own environment.

Anatomy of a ComfyUI Workflow

Workflows are constructed from a network of connected nodes. Each node performs a specific function within the generation process, while the connections dictate the flow of data between them.

A standard text-to-image workflow typically includes nodes for loading the model, processing the prompt, initializing image data, generating the image, decoding the result, and saving the final file.

• Model loader: Retrieves the model required for generation.
• Text encoder: Translates your prompt into a format the model can interpret.
• Sampler: Executes the generation process based on the chosen settings.
• VAE: Facilitates the conversion between latent data and visual image data.
• Save Image: Writes the completed image to your disk.

There is no requirement to build every workflow from the ground up. ComfyUI offers pre-built workflow templates, and a vast array of community-created workflows can be downloaded and opened directly.

Loading Existing Workflows

Starting with an existing workflow is often the most efficient method. ComfyUI includes example workflows for various models and tasks, and community platforms offer even more extensive options.

Often, workflow data is embedded within the metadata of workflow images. You can drag the image directly into ComfyUI or utilize Workflows \u2192 Open to load it. The workflow will then appear on the canvas with all nodes and settings pre-configured.

After loading, verify which models the workflow requires. If any files are missing, ComfyUI can identify absent models for supported templates. For other workflows, you may need to manually locate and install the required models.

Sourcing Models for ComfyUI

Models are commonly available on repositories like Hugging Face and Civitai, as well as on the official project pages for specific models. The critical step is ensuring the model is compatible with your chosen workflow.

Do not assume that any model file will function within any workflow. Different model architectures often necessitate specific loaders and supporting files.

Before downloading a model, verify the following:

• The model's architecture and version
• The compatible ComfyUI workflow
• The model file format
• Recommended VRAM and hardware specifications
• Any necessary VAE, text encoder, LoRA, or other auxiliary files
• The model's licensing and usage restrictions

ComfyUI accommodates various model file types, with their storage location depending on the model category. For instance, checkpoints are stored in models/checkpoints, LoRAs in models/loras, and VAEs in models/vae. Newer models may utilize directories such as models/diffusion_models and models/text_encoders.

Installing Models

Once you have downloaded a model, place it in the directory expected by the workflow. You can then select it within the corresponding model loader.

For instance, a checkpoint might be located in:

ComfyUI/models/checkpoints/

Conversely, a LoRA might be found in:

ComfyUI/models/loras/

If the newly installed model does not appear in the list, try refreshing the interface or restarting ComfyUI.

Installing Custom Nodes

Many advanced ComfyUI workflows rely on custom nodes that are not part of the standard installation. If these dependencies are missing, the workflow may indicate missing nodes.

ComfyUI includes a Manager to facilitate the installation of custom nodes. Alternatively, you can install nodes manually by placing their repositories in the custom_nodes directory and resolving any required dependencies.

Exercise caution and only install custom nodes from trusted sources. Since custom nodes can contain executable code, they may introduce their own dependencies and security considerations.

Executing and Modifying Your Workflow

With the necessary models and custom nodes in place, review the key settings within your workflow. Begin by checking the model, prompt, image dimensions, and sampling parameters.

When everything is configured correctly, use the Queue button to initiate the workflow. ComfyUI will process each step and generate the output as defined by the workflow structure.

Afterwards, you can modify specific elements of the workflow without reconstructing the entire graph. You can add LoRAs, connect input images, switch samplers, apply upscalers, or adjust other settings to refine the result.

Saving Your Workflows

Save any workflow you plan to reuse. While a workflow contains the node graph and its settings, it does not inherently include the model files themselves. It is crucial to keep track of which models and custom nodes the workflow depends on.

This is particularly significant when transferring a workflow to another computer or cloud desktop. You will likely need to install the same models and custom nodes before the workflow will function correctly.

Explore ComfyUI on DaDesktop

Running ComfyUI does not require a new GPU purchase. If your current machine lacks sufficient GPU resources, you can execute ComfyUI on a cloud desktop, utilizing it on an as-needed basis.

DaDesktop offers cloud desktops equipped with dedicated GPU resources, ideal for workloads such as AI image and video generation. You can install ComfyUI, download your preferred models, and develop your own workflows without the need to add a dedicated GPU to your local setup.

Discover more about AI image and video generation on DaDesktop. You can also view the available GPUs and select a configuration that suits your specific model and workflow requirements.