ComfyUI Advanced Inpainting Workflow with Flux Redus

Hi friends! Welcome back to my Blog, and Happy New Year! 🎉 I hope you’re all doing great. Today, I’m sharing something super cool—how to use Flux Redux to combine two pictures into one. Don’t worry, it’s easier than it sounds, and I’ll explain everything step by step.

Let’s say you have a picture, but you want to change part of it or add something new. For example, imagine you have a photo of a girl standing outside, and you want her to look like Spider-Woman instead. Normally, this can get tricky because the result might not look right. But with Flux Redux, it’s like magic—it makes everything look perfect, like it was always supposed to be that way.

In simple words, inpainting is fixing or changing parts of an image. It’s like when you erase part of a drawing and draw something new in its place. With Flux Redux, it gets even better because you can use another picture as a style reference to make it all match.

How to Get Started

Step 1: Download the Tools

The workflow has three parts:

First, go to comfyuiblog.com and download the inpainting workflow. After you’ve downloaded it, open ComfyUI on your computer and load the workflow.

1. Auto Prompt Inpaint + Style Transfer: The computer figures out what to do automatically.
2. Manual Text Prompt Inpaint: You type what you want the picture to look like.
3. Inpaint Without Flux Redux: A simple way to change the picture without using styles.

You’ll need some extra files called models. Here’s what to do:

1. Diffusion Model:
   • Download Flux Dev 1 Fill and save it in the “diffusion model” folder.
   • If your computer isn’t super fast, use Flux 1 Fill Dev FP8 instead.
2. Dual Clip Loader:
   • Choose T5XXL FP16 for the first option.
   • For the second option, pick Clip I Flux and set it to “Flux.”
3. Style Model:
   • Download this file and put it in the “Style Model” folder.
4. Clip Vision Model:
   • Get the SigClip Vision Patch 14 384 file and save it in the “Clip Vision” folder.

1. Upload Your Pictures:
   • First, upload the image you want to change (for example, a girl standing outside).
   • Then, upload the second image you want to use as the style (like Spider-Woman).
2. Mask the Area:
   • Cover the part of the picture you want to change. In this case, mask the girl.
3. Add a Text Prompt:
   • Write what you want, like “Spider-Woman.”
4. Generate the Image:
   • Click the button to create your new picture.

Without Flux Redux, the results can be a little messy. The colors might not match, or the picture might look fake. But with Flux Redux, it’s like magic! For example:

• The new Spider-Woman keeps the same pose and expression as the original girl.
• The background stays exactly the same, with all the same lighting and details.

• Be careful with the mask: If you mask the area properly, you’ll get better results.
• Try different styles: Upload different pictures to see what works best.
• Use the right model: If your computer isn’t super fast, go for the FP8 version of the model.

• FLUX.1-Fill-dev_fp8: FLUX.1-Fill-dev_fp8.safetensors · 1038lab/FLUX.1-Fill-dev_fp8 at main
• FLUX.1-Fill-dev: flux1-fill-dev.safetensors · black-forest-labs/FLUX.1-Fill-dev at main
• Flux1 Redus: flux1-redux-dev.safetensors · black-forest-labs/FLUX.1-Redux-dev at main
• Clip Vision: model.safetensors · google/siglip-so400m-patch14-384 at main