Introducing loop unleashed, which allows you to easily create deepfake images and videos

overview

Introducing a WEB UI library that allows you to easily replace faces in still images and videos. Originally there was a library called roop, but a fork of it is roop-unleashed. It has been announced that the original roop will no longer be updated. It seems like he has lost the motivation to do it.

GitHub - C0untFloyd/roop-unleashed: Evolved Fork of roop with Web Server and lots of additions

preparation

git clone https://github.com/C0untFloyd/roop-unleashed.git
cd roop-unleashed
pip install pip install -r requirements.txt

execution

python run.py

How to use

Insert the “photo of the face image you want to use” in 1. If there are multiple faces, click 3 on any face.
Insert the target image or video into 2. For videos, specify the output frame rate as 4. If it is 0, it is the same as the original.

5 to Codeformer. Anything other than None is fine. If you leave it as None, the converted face will be of low quality and unnatural. The number below is a parameter that determines how sharp you want it to be. A value higher than the default would be good.
6 is the setting to change whose face in the target image.
I don’t have to do anything else. Each function has the same meaning as in English.

Start conversion with start button

Advice for videos

Even in videos, the face changes quite naturally. Conversion speed highly depends on machine specs. If you think it’s too slow, try reducing the Video FPS. I can’t use the NVIDIA GPU on my Mac, so a 5-minute video was the limit of my patience. Maybe it would be a little faster if I installed NVIDIA?

Other features

I haven’t tried it, but it seems like you can convert in real time using the live camera.
I think you should leave the settings as default.

Images used

https://unsplash.com