ComfyUI - Getting Started : Episode 1 - Better than AUTO1111 for Stable Diffusion AI Art generation

Scott Detweiler
13 Jul 202319:01

TLDRIn this video, Scott Weller introduces Comfy UI, a versatile tool for AI art generation that surpasses Automatic 1111 in capabilities. As the head of quality assurance at, Scott shares his expertise and daily experience with the tool, guiding viewers through its installation and basic workflow. He demonstrates how to create and refine AI-generated images using prompts, models, and samplers, highlighting Comfy UI's advanced features and potential for complex, creative projects. Scott also mentions upcoming videos and his consideration of starting a podcast to share more insights.


  • 🎨 Introduction to Comfy UI, considered the best tool for AI art generation at the moment.
  • πŸ‘¨β€πŸ’Ό Scott Weller, head of quality assurance at, shares his experience with Comfy UI daily.
  • πŸ”§ Comfy UI can handle a wide range of tasks from control nets to training models.
  • πŸ’» The tool requires a computer with more than 3GB of video RAM, but can also run on a CPU, albeit slower.
  • πŸ”— A link will be provided for viewers to download and install Comfy UI easily.
  • πŸ› οΈ The video is the first in a series that will familiarize users with the process and workflow of Comfy UI.
  • 🎭 The tool's capabilities surpass those of other products like Automatic 1111, offering more customization and flexibility.
  • 🌟 Tips and tricks for building a graph from scratch will be shared, highlighting the workflow for creating AI art.
  • πŸ”„ Demonstration of how to add, duplicate, and manage nodes within the Comfy UI platform.
  • πŸ–ŒοΈ Explanation of the different steps involved in creating an image, from prompts to sampling and encoding.
  • πŸš€ Upcoming release of SD Excel and how it will complement the capabilities of Comfy UI.

Q & A

  • What is the main topic of the video?

    -The main topic of the video is an introduction to Comfy UI, a tool for AI art generation, and its capabilities compared to other tools like Automatic 1111.

  • Who is the speaker of the video?

    -The speaker of the video is Scott Weller, who works as the head of quality assurance at Stability AI.

  • What are some of the features that Comfy UI offers for AI art generation?

    -Comfy UI offers features such as adding nodes to the graph, using different models like Nets and Loris, applying prompts, sampling, and encoding images, as well as upscaling and fine-tuning the generated art.

  • How does the speaker describe the workflow with Comfy UI?

    -The speaker describes the workflow with Comfy UI as starting with adding a node to the graph, selecting a model, applying prompts, sampling, and encoding the image back into a visual format. It involves a series of steps that can be customized and expanded upon.

  • What are the system requirements for running Comfy UI?

    -Comfy UI can run on almost every computer with more than three gigabytes of video RAM. It can also work on a computer with just a CPU, although it will be slower.

  • How does the speaker compare Comfy UI to Automatic 1111?

    -The speaker compares Comfy UI to Automatic 1111 by stating that while both can generate AI art, Comfy UI offers more flexibility and capabilities, allowing for a wider range of customization and more complex workflows.

  • What is the significance of the 'sampler' in the workflow?

    -The 'sampler' is significant in the workflow as it is used to sample from the model based on the prompts provided. It helps in creating the initial version of the AI-generated image, which can then be refined further in the process.

  • How does the speaker suggest organizing the nodes in the Comfy UI graph?

    -The speaker suggests organizing the nodes by labeling them with clear titles, changing colors for visual distinction, and using the collapse feature to minimize less important or non-tweakable nodes. This helps in managing complex workflows and keeping track of different components.

  • What is the purpose of the 'latent image' in the workflow?

    -The 'latent image' serves as a noise-filled starting point that is fed into the sampler along with the model and prompts. It is used to generate the initial version of the AI art, which can then be refined and denoised to create the final image.

  • How does the speaker plan to continue the series of videos?

    -The speaker plans to continue the series of videos by exploring more advanced features and techniques with Comfy UI, demonstrating different models, and possibly starting a podcast to share more insights and information.

  • What is the speaker's recommendation for users who want to get started with Comfy UI?

    -The speaker recommends users to install Comfy UI and start exploring its features. He suggests that users can begin with the basics and then expand their workflow as they become more familiar with the tool and as new models are released.



🎨 Introduction to Comfy UI for AI Art Generation

In this introductory segment, Scott Weller highlights Comfy UI as the premier tool for AI art generation, surpassing alternatives like Automatic 1111. As the head of quality assurance at, Scott shares his daily experience with the tool, emphasizing its versatility in handling various AI art tasks, from control nets to training models. He provides a link for viewers to install Comfy UI, noting its compatibility with most computers, even those with just CPU capabilities. Scott sets the stage for a series of videos to familiarize users with the workflow and capabilities of Comfy UI, especially in anticipation of the upcoming release of SD Excel.


πŸ› οΈ Building a Workflow from Scratch

Scott Weller delves into the practical aspects of using Comfy UI by demonstrating how to build a workflow from scratch. He explains the process of adding nodes to the graph, starting with a loader to load a checkpoint, which represents the model. He then discusses the importance of organizing the workflow by using different colors and labels for positive and negative prompts. Scott also covers various methods for duplicating nodes and maintaining a clean graph for ease of use, especially when dealing with complex workflows. He emphasizes the customizability and the continuous addition of new nodes to the product.


🌐 Upscaling and Sampling with Advanced Techniques

This paragraph focuses on advanced techniques within Comfy UI, such as upscaling and resampling. Scott introduces the concept of a latent image, which is the noise fed into the sampler to create the final image. He explains the process of using a sampler with a model and emphasizes the importance of setting the right parameters, like the number of steps and the denoise level. Scott also demonstrates how to upscale the latent image by two times and how to use different samplers at different stages of the process. He introduces the advanced case sampler for more control over the sampling process and shows how to neaten up the graph using reroute nodes.


πŸŽ₯ Final Touches and Future Tutorials

In the concluding segment, Scott Weller wraps up the tutorial by discussing the final steps in the workflow, including saving and previewing the generated images. He talks about the ability to drag images directly into the graph for further processing and the convenience of having a unified seed for all noise generation. Scott also mentions the potential of using Comfy UI with new models and the smart features it offers, like automatic GPU detection and CPU-only mode. He invites feedback from viewers and expresses his intention to create more content, including a podcast, to share valuable insights and keep the community updated on the latest developments in AI art generation.



πŸ’‘Comfy UI

Comfy UI is described as an advanced and powerful tool for AI art generation. It is the main subject of the video, with the speaker, Scott Weller, highlighting its capabilities and ease of use. The tool is favored over other AI art generators like 'automatic 1111' due to its versatility and extensive features, allowing users to create, modify, and refine AI-generated images through a graphical interface.

πŸ’‘AI Art Generation

AI Art Generation is the process of creating visual art using artificial intelligence. In the context of the video, it refers to the use of AI tools like Comfy UI to generate images based on user inputs, such as prompts and model selections. The speaker discusses the intricacies of this process, including the use of different models, prompts, and sampling techniques to refine the generated art.


Checkpoints, in the context of the video, refer to saved states of AI models that can be loaded into Comfy UI for further manipulation or generation of art. They are essential for continuing work from a previous point or reusing models in new projects.


Prompts are inputs or instructions given to AI models to guide the generation of specific images or art. In the video, positive and negative prompts are used to refine the AI's output, with the positive prompt being the desired outcome and the negative prompt serving as an antithesis to guide the AI away from undesired features.


Sampling in the context of AI art generation refers to the process of taking small parts or 'samples' of the AI's output at different stages of the generation process to refine the final result. It involves selecting portions of the generated image that align with the desired outcome and further processing those parts to improve the overall quality.

πŸ’‘Latent Images

Latent images are representations of the initial, unprocessed data or noise that AI art generation tools use as a starting point for creating images. They are then refined and denoised through the application of prompts and models to produce the final visual output.


An autoencoder is a type of artificial neural network used for unsupervised learning of efficient codings. In the context of the video, it is used as part of the AI art generation process to encode and decode the latent images, transforming them into the final visual output.

πŸ’‘Quality Assurance

Quality assurance (QA) refers to the process of ensuring that a product or service meets certain standards of quality. In the video, Scott Weller mentions his role in quality assurance at Stability AI, which implies that he is responsible for testing and verifying the performance of AI tools like Comfy UI.


Workflow in the context of the video refers to the sequence of steps and processes followed to generate AI art using Comfy UI. It involves the use of nodes, prompts, models, and other tools within the interface to create and refine the desired images.


Upscaling in the video refers to the process of increasing the size or resolution of an image or latent data. This is done to improve the quality or detail of the AI-generated art, allowing for larger, more detailed outputs.

πŸ’‘Stability AI

Stability AI is the company mentioned in the video that works with the speaker and is responsible for the development of Comfy UI. The company is involved in the creation and improvement of AI tools for art generation.


Introduction to Comfy UI, considered the best tool for AI art generation at the moment.

Comfy UI can do everything that Automatic 1111 does and more, including control nets, Loris, and training models.

The presenter works at as the head of quality assurance and uses Comfy UI daily.

Comfy UI is accessible via a simple git installation and works on computers with more than 3GB of video RAM, even on those with just CPU albeit slower.

Comfy UI allows for building a workflow from scratch, providing tips and tricks for the process.

The tool offers various options for adding nodes to the graph, such as right-clicking, double-clicking, or using clipboard functions.

Custom nodes will be added to the product over time, enhancing its capabilities.

The importance of labeling nodes with colors and titles for clarity in complex workflows.

The process of using prompts, samplers, and models to generate AI art, including the use of positive and negative prompts.

The ability to upscale latent images and adjust settings like steps, seed, and denoise for better image quality.

The use of different samplers, such as Euler and Keras, and their impact on the step process.

The convenience of collapsing and expanding nodes for a cleaner workflow.

The capability to save and preview images directly within the UI.

The potential for creating advanced workflows and the flexibility to adapt with new models.

The presenter's intention to continue making videos and potentially start a podcast to share more information.

The ability to drag images created in Comfy UI directly into the graph for further processing or queuing.