Is This The Best Tool For Deepfakes? Faceswap SDXL

This new Face Swap plugin lets you put your face onto any image. It doesn't need any training at all. It just takes the open sourced insightface architecture and packages it into an extension in stable diffusion. Show your works in the comments.
Disclaimer: This video is created solely for educational purposes. This video in no way promotes creating deepfakes of real people. This technique should never be used for creating deepfakes. This is mainly for storyboards and graphic novels with consistent characters.
Download Faceswap Extension: github.com/glu...
Install Refiner extension for SDXL: • SDXL Refiner Extension...
My other contents at x.com/DebarghaDB
DM me at / debargha_bhattacharjee
Drop an email to [e-mail]

[@justthecoolestmusicjohne6105]

Can't understand a word you are saying.

[@doyen86]

When i use this no matter the image or the quality of the image, my faceswaps always come out blurry. Any idea on how to fix this?

[@Debargha_Bhattacharjee]

There might be some error in the settings. Can you try deleting the extension and reinstalling it? It might help.

[@AnilSari1903]

How can we upload existing images and swap there the faces? Is this only used for live generated images?

[@Debargha_Bhattacharjee]

Watch this video from 01:55

[@Comic_Book_Creator]

thanks, can you do more with Comfyui:

[@Debargha_Bhattacharjee]

Sure, I'll try to.

[@Stick3x]

Even though you speak too fast I am subscribing. I also love your accent. Thank you for this tutorial.

[@Debargha_Bhattacharjee]

Thank you and welcome to the channel.

[@iplt20-20official]

Where can I download face safesensor model? I searched but I can't find

[@Debargha_Bhattacharjee]

github.com/glucauze/sd-webui-faceswaplab?fbclid=IwAR2yz60PO8LMTU370928Okqp-zvCzb1a0NiWGAmkTdsqylStpwI0JBsb9H0 here you go

[@mr.entezaee]

I didn't find any safesensor model there@@Debargha_Bhattacharjee

[@walidflux]

i usually train lora for faces but this is good to add after generating with Lora or before to help lora in img2img and inpainting

[@Debargha_Bhattacharjee]

This completely replaces the face. Therefore using this before or after using the lora won't bring any change as per my knowledge.
Although if you use it after using lora then that would help mitigate some problems like sometimes we get slightly distorted images in some cases. But in my opinion, if you use this and try to match the pose and angle of the generated image with the input image, you won't need the lora at all.

[@gordonbrinkmann]

I've switched to ComfyUI a while ago now because I like the nodes workflow which I am used to from other software and I can generally create larger images with my hardware when using ComfyUI instead of A1111. So it would be really cool if there was some kind of implementation of FaceSwap into ComfyUI as well...

[@Debargha_Bhattacharjee]

I shall try to look up if there is any way to integrate this into comfyui. If there is one, that would be great for me too. Although I don't prefer the node based workflow much, but comfy is the way to go as new developments of sdxl are based on comfy.
A roundabout way would be to use the comfyui extension for automatic 1111. That would get the job done.

[@gordonbrinkmann]

@@Debargha_Bhattacharjee I actually didn't know about the comfy extension for A1111. But if it only allows you to use nodes I don't think I can use it. My problem with A1111 nowadays since SDXL came out is that with my hardware I can hardly render images while with ComfyUI it's still quick and I get large images.

[@Debargha_Bhattacharjee]

It was the same for me. I waited for sdxl official release to get my hands on it jut because I did not want to use comfyui. But seeing how slow and limited it is automatic1111, I finally decided to take the plunge and it's been great ever since. Yeah, little bit complicated, but great in case of results and tweaking capabilities.@@gordonbrinkmann

[@godpunisher]

you may use ReActor Node for ComfyUI.