Install Stable Diffusion 3 Locally: Step-by-Step with StableSwarmUI & ComfyUI

The video tutorial begins by introducing the release of Stable Diffusion 3 and the two interfaces available for use: Swarm UI and Comfy UI. The speaker opts to demonstrate the installation process on Windows, highlighting the prerequisites for Windows 10 and the automatic process for Windows 11. The audience is guided through downloading a batch file, setting up a new folder, and running the installer, which includes downloading .NET 8 and configuring Git. The tutorial also covers the Swarm UI's installation, customization options, and legal notices regarding the MIT license and personal use limitations.

The script continues with instructions on how to access and download Stable Diffusion 3 models from the Hugging Face website. It details the process of creating an account, verifying email, and agreeing to terms for non-commercial use. The speaker then explains how to select and download specific models, such as the 'sd3 medium' version, and how to integrate these models into the Swarm UI by using the refresh button and model selection options. Additionally, the video demonstrates how to generate images using various settings, resolution options, and prompts, emphasizing the first-time download of necessary files and the use of text encoders.

This paragraph delves into the customization of model settings within the Swarm UI, discussing recommended configurations for different models like Juggernaut and sd3. It explains how to adjust steps, CFG scale, and sampling settings for optimal image generation. The tutorial also touches on the use of text encodings and the impact of RAM usage, suggesting the selection of appropriate options based on the capabilities of the user's video card. The process of saving generated images and accessing them in the output folder is also covered, along with the option to create shortcuts for easier access.

The script shifts focus to the installation of Comfy UI, providing a step-by-step guide for downloading, extracting, and launching the interface. It mentions the need for an Nvidia GPU and the use of a specific batch file to run the interface. The tutorial also introduces the Comfy UI manager for installing custom nodes, updating models, and managing the UI. The speaker demonstrates how to create a shortcut for Comfy UI, change its icon, and use it to launch the interface, highlighting the additional manager button for accessing advanced options.

This section outlines the process of importing Stable Diffusion 3 models into Comfy UI, including the placement of models in the appropriate folders and the use of the manager to load and test models. The script discusses the use of workflows, the installation of missing nodes, and the customization of workflow settings for different models. It also shows how to save and organize workflows, troubleshoot node errors, and explore the workflow gallery for additional options. The tutorial emphasizes the importance of checking model compatibility and the potential for conflicts with custom nodes.

The final paragraph demonstrates the testing of various models and workflows within Comfy UI, showcasing the generation of images using different prompts and settings. It highlights the ability to save generated images and the process of troubleshooting node errors by installing missing components. The script also encourages exploring the workflow gallery and customizing settings for optimal results. The tutorial concludes with an invitation to join a Discord server for further discussions and sharing of AI-generated images, and mentions an additional YouTube channel focused on AI experiments.

In the concluding paragraph, the speaker thanks the audience for watching and encourages them to like the video if they found it useful. A piece of background music is also noted, signaling the end of the tutorial.