ComfyUI : NEW Official ControlNet Models are released! Here is my tutorial on how to use them.

Scott Detweiler
20 Aug 202315:59

TLDRThe video introduces the release of official control net models for the sdxl platform, emphasizing their efficiency and versatility. The host guides viewers through the installation of a manager for node management, the integration of models from the hugging face repository, and the use of preprocessors. The demonstration showcases the application of control net models in creating detailed and intricate images, highlighting the customization options and the creative potential of the technology.


  • 🚀 Introduction of the official control net models for the community.
  • 🛠️ Importance of installing the manager for handling custom nodes efficiently.
  • 🔄 Correcting the mistake from a previous video, emphasizing the use of 'git clone' instead of 'get clone'.
  • 📱 Navigating to the GitHub repository to clone the manager for local installation.
  • 🔧 Utilizing the manager to install custom nodes and preprocessors from the sdxl official Hugging Face repository.
  • 🔍 Explanation of the two types of control net preprocessors and the recommendation to use the work in progress version.
  • 🧠 Understanding the difference between normal maps and depth maps, and their applications in the creative process.
  • 🖼️ Demonstration of how to use the candy edge detector and depth map preprocessors to enhance images for control net input.
  • 🔄 Discussing the installation of sdxl models, also known as control lora's, from the Hugging Face repository.
  • 🎨 Walking through the process of setting up the control net in the node system, including selecting appropriate models based on system memory.
  • 📸 Using a depth map as a conditioning element in the creative process, allowing for a blend of the original image and desired outcome.

Q & A

  • What is the main topic of the video?

    -The main topic of the video is the introduction and usage of the official control net models for the Comfy platform.

  • What is the first step in using the control net models?

    -The first step is to install the manager for Comfy, which is highly recommended for managing custom nodes.

  • Where can the manager for Comfy be installed from?

    -The manager can be installed from a GitHub repository, with the link provided in the video description.

  • How do you install custom nodes using the Comfy manager?

    -You can install custom nodes by going to the Comfy local installation, navigating to custom nodes, and using the command to clone the desired node from GitHub.

  • What is the purpose of the control net preprocessors?

    -The control net preprocessors are used to process images before they are used in the control net models, enhancing features like edges or depth maps to guide the generation process.

  • How can you install the sdxl models?

    -The sdxl models can be installed from the Hugging Face repository, and then added to the Comfy installation under the control net folder.

  • What is the significance of the 'control net' and 'control net apply advanced' options in Comfy?

    -The 'control net' and 'control net apply advanced' options in Comfy allow users to apply the control net models to their images with more customization and control over the generation process.

  • How do you use a depth map in the control net model?

    -A depth map can be used in the control net model by loading it as a preprocessed image and then applying it to the control net apply advanced option, which will take into account the depth information during the generation process.

  • What is the role of the latent node in the process?

    -The latent node represents the random noise vector that is used as the starting point for the generation process. It needs to be a specific size (1024 or larger for sdxl) and is fed into the model for generation.

  • How can you control the influence of the control net on the generated image?

    -The influence of the control net can be controlled by adjusting the strength settings, as well as the start and end points where the control net's influence is most applied during the generation process.

  • What is the purpose of the 'encoder' nodes in the script?

    -The encoder nodes are used to process the positive and negative prompts, which guide the generation process by providing the model with desired and undesired characteristics for the output.



🚀 Introduction to Control Net Models and Setup

The speaker, Scotty, introduces the availability of official Control Net models and outlines the process for setting them up. He emphasizes the importance of installing a manager for handling custom nodes, which simplifies the process significantly. Scotty corrects a previous mistake regarding the installation process and provides a quick guide on installing the manager from a GitHub repository. The video's focus is on using the Control Net models rather than installation, and Scotty mentions the need for preprocessors, which will be covered later in the video.


🛠️ Exploring Preprocessors and Control Net Functionality

Scotty delves into the functionality of preprocessors, demonstrating how they can enhance images by extracting context and combining elements to create new visuals. He discusses the use of the Candy Edge detector and depth maps for refining image outlines and details. The speaker also explains the distinction between normal maps and depth maps, highlighting their different applications. Scotty then illustrates the application of Control Net models by loading an image and using different preprocessors to modify and enhance it according to specific requirements.


🎨 Applying Control Net Models and Preprocessors

In this section, Scotty focuses on applying Control Net models and preprocessors within the software. He explains how to load positive and negative encoders into the Control Net and emphasizes the importance of using the correct model and image for each step. Scotty also discusses the possibility of chaining multiple Control Nets together for more complex image processing. He provides a brief overview of the settings and options available within the software, such as the K sampler and VAE, and how they can be adjusted to achieve desired outcomes.


🌟 Finalizing the Image and Prompt Settings

Scotty concludes the video by discussing the final steps in processing the image using Control Net models. He explains how to adjust the strength and focus of the model's application throughout the image creation process. The speaker also demonstrates how to fine-tune the image by controlling the adherence to the depth map at different stages of the process. Scotty provides a prompt example and shows the resulting image, highlighting the effectiveness of the Control Net models and preprocessors in achieving the desired visual outcome. He wraps up by thanking the viewers and the supporters of the channel.



💡sdxl official control net models

The 'sdxl official control net models' refers to a set of newly released machine learning models that are central to the video's content. These models are designed to be used within a specific software or framework, providing users with advanced capabilities for image processing and generation. In the context of the video, these models are being introduced as a powerful tool for users to create and manipulate images in a more controlled and detailed manner.


The 'manager' in the video is a software tool or plugin that simplifies the process of managing and installing custom nodes or extensions within the Comforty (Comfy) application. It is highly recommended for users to install this manager to facilitate the use of control net models and other necessary components. The manager streamlines the process of accessing and utilizing the newly released sdxl models, making it easier for users to integrate them into their workflows.

💡Hugging Face repository

The 'Hugging Face repository' is an online storage location, typically on a platform like GitHub, where the sdxl models are stored and made accessible to the public. This repository serves as a central hub from which users can download and install the control net models for use in their projects. It is an essential resource for obtaining the latest models and staying up-to-date with the developments in the field.


In the context of the video, 'preprocessors' are specialized tools or algorithms that prepare and modify input data, such as images, before it is processed by the main model. For instance, an edge detector or a depth map generator would be a type of preprocessor. These preprocessors enhance the input data to ensure that it is in the correct format and contains the necessary features for the subsequent model to effectively generate or transform the image.

💡control net

A 'control net' refers to a specific type of neural network architecture used in the video for image processing tasks. It is designed to take in certain types of preprocessed data and generate outputs that can be controlled or manipulated to achieve specific visual effects. The control net is a key component in the creative process described in the video, allowing users to guide the generation of images based on their preferences and requirements.

💡custom nodes

In the context of the video, 'custom nodes' are individual components or building blocks within the Comforty (Comfy) software that can be installed and used to extend its functionality. These nodes represent various features or tools that users can integrate into their projects to perform specific tasks, such as image processing or model management. Custom nodes enhance the capabilities of the software, allowing users to create more complex and sophisticated projects.


A 'workflow' in the video refers to a sequence of steps or processes that are followed to achieve a specific outcome, particularly in the context of image generation and manipulation. It involves the use of various tools, models, and techniques to transform input data into the desired output. The workflow is crucial for users to understand, as it guides them through the process of using the control net models and preprocessors to create images that meet their specifications.

💡depth map

A 'depth map' is a type of image or representation that encodes the distance of objects from the camera or viewer. It is a grayscale image where lighter areas indicate objects closer to the viewer, and darker areas indicate objects further away. Depth maps are used in image processing and computer graphics to create a sense of depth and three-dimensionality in visual content. In the video, depth maps are utilized as part of the preprocessing step to add depth information to images before they are processed by the control net models.


In the context of the video, a 'prompt' is a text input or instruction given to the AI model to guide the generation or transformation of an image. The prompt serves as a creative directive, telling the model what kind of image the user wants to create. It is a critical part of the process, as it influences the final output and helps the model understand the user's intent.


In the field of machine learning and AI, a 'latent' refers to a hidden or underlying variable that is not directly observable but can be inferred from the data. In the context of the video, a 'latent' is a numerical representation or vector that captures the essential features of an image or data set. This latent representation is used as input for generative models to create new images or modify existing ones based on the provided prompts and control settings.


Introduction of the official control net models for the sdxl platform, marking a significant update for users.

Recommendation to install the manager for efficient handling of custom nodes, streamlining the process of using the new models.

Clarification on the correct method to install the manager, correcting a previous mistake in a video tutorial.

Explanation of the process to acquire models from the sdxl official Hugging Face repository.

Importance of preprocessors in the workflow and how to install them for optimal use.

Demonstration of the manager's capability to simplify the installation of custom nodes, showcasing its user-friendly interface.

Insight into the use of control net preprocessors and their role in enhancing the creative process.

Discussion on the architectural implementation of control Lora's, emphasizing their efficiency and compact design.

Practical guide on installing sdxl models from Hugging Face, including the use of the control Lora's and their respective folders.

Explanation of the advantage of using a single location for model storage when working with both Comfy and Automatic 1111.

Showcase of the preprocessors' functionality, including the candy edge detector and depth map, illustrating their impact on image processing.

Discussion on the combination of different control nets for enhanced results, such as using both candy and depth for detailed and accurate image representation.

Walkthrough of the process to load and apply the control net models within the Comfy interface, including the selection of appropriate versions based on system memory.

Explanation of the conditioning aspect of the control net process, highlighting its significance in shaping the final output.

Demonstration of the control net's ability to be stacked and chained for complex and intricate image manipulation.

In-depth look at the settings and parameters involved in the control net application, such as strength, start, and end points, offering users greater control over the creative process.

Conclusion and call to action for viewers to experiment with the new models and share their thoughts, fostering a community of engaged and creative users.