The video introduces FLUX tools, including inpainting, Redux, Depth, and Canny, and provides a step-by-step guide on how to run these models locally using ComfyUI. The presenter demonstrates the setup process, shares specific settings for optimal results, and encourages viewers to experiment with the tools to achieve diverse and impressive image outputs.
In the video, the presenter introduces the new FLUX tools, which include features such as inpainting, Redux, Depth, and Canny. The video aims to guide viewers on how to run these tools locally. The presenter begins by exploring the FLUX website, where users can read about the different models and their functionalities. The inpainting model allows users to fill in missing parts of images, while the Depth model applies depth information to images. The Redux model generates variations of input images, showcasing the versatility of the FLUX tools.
To get started, the presenter explains the process of downloading the necessary Loras (models) for Depth and Canny. Viewers are instructed to agree to the terms before downloading and placing the models into the appropriate folders. The video emphasizes the importance of using the updated version of ComfyUI, a user interface for running these models. After downloading a sample shark image, viewers are guided through the workflow setup in ComfyUI, including selecting the correct models and adjusting settings for optimal results.
The presenter shares their specific settings for the Depth model, including the use of the Flux one depth FB8 model and adjustments to the K sampler settings. They demonstrate how to preprocess images for depth input and showcase the resulting images, highlighting the improvements made through upscaling. The video also covers the Canny model, where similar steps are followed, but with adjustments to the Lora settings for better results. The presenter emphasizes the importance of experimenting with settings to achieve desired outcomes.
Next, the video transitions to the Redux process, where viewers learn to download and set up the Redux model and the necessary clip file. The presenter explains how to load the models and images into ComfyUI without needing to modify the workflow significantly. By pressing the render button, users can create variations of their images, showcasing the effectiveness of the Redux model in generating diverse outputs while retaining details from the original image.
Finally, the video discusses the fill-in painting model, which requires downloading a larger model file. The presenter demonstrates how to use the mask editor to select areas for inpainting and run the model to achieve the desired results. The video concludes by praising the FLUX tools for their user-friendly workflows and impressive capabilities, encouraging viewers to explore these models and share their experiences. The presenter thanks the audience for watching and invites them to leave feedback.