ReActor Face Swapping of images and Videos in A1111

#stablediffusion #reactor #faceswap #faceswapping #a1111 #aivideo #aiimages

00:00:00 introducing ReActor in A1111 and faceswapping showcases
showcases in videos and image face swapping
00:00:18 why ReActor instead of Roop?
00:01:16 installation of Reactor
00:02:18 using ReActor in A1111
show settings and example usage
00:04:51 face swapping in batch image for videos
show how to generate frames from Davinci resolve and swap faces in a video using ReActor frame by frame in A1111

Computer Specs:
Laptop: Legion 5 Pro
Processor :AMD Ryzen 7 5800H , 3201 Mhz
System RAM: 16.0 GB
Graphics GPU: NVIDIA GeForce RTX 3070 Laptop GPU 8GB

ReActor extension for Stable Diffusion

faceswaplab (another extension that you maybe interested in)

A1111 for stable diffusion.

This video is about face swapping in Automatic 1111 using ReActor which is a fast 1 image face swap extension built on top of Roop that I have explained in a previous video.
We will see how it’s done in text2img, img2img, and face swapping in a video.
o Now Why ReActor instead of Roop?
o ReActor is allows uncensored NSFW images unlike the original Roop
o More features, gender detection, better scaling
o Allows to use multiple faces is source image to swap multiple face in target images
o Roop project has ceased development.
o Installing ReActor for A1111
o Using ReActor with text2image
o Using it with img2img
o Using it for face swapping in videos
o Note: After detailer can be used for face swap in images too and with LoRA may give very good results if LoRA of a person is present.

Face swapping is very useful to create consistent faces as well.

Notes: it is better if source and target images to have similar face structure, that will give better results.
So as a conclusion, ReActor is more flexible and gives us more control in comparison to roop and is a fast and effective face swapper extension for Automatic 1111.

Attribution:
Honye Sanges for pexels-los-muertos-crew-8447488 (720p) video
and video id ID_4723042

Thanks to Insightface developers as well who have provided the underlying technology for these amazing tools and their face detection systems, check more at:

Note: ReActor can be used from Extra's tab and batch processing which can have some advantages if you dont want to go through img2img tab.

[ElDespertar]

Thank you so much for this super useful tutorial!

[chrissinclair]

Thanks for this I finally know how to faceswap now haha

[KM-zxzz]

You are so wonderful for sharing such great information!
It was very useful for me because there is not much information available online in Japan.
Thanks to you, I have been able to try various quality video generation.

I have two questions I would like to ask.

In roop, I was able to change only the face by setting the Denoising strength value to 0, but in Reactor, the hair and the background around the face were also affected a little.

Also, the images generated by the batch process are not consistent when compared before and after, and the images appear to flicker due to slight differences.

Do you know how to solve these problems?

[chilecayenne]

Thank you for the interesting video! Please forgive an extreme noob question.
But the first step here had you downloading and installing MS Visual Studio. Does this mean ReActor will only work on MS Windows systems and not say on Linux or OSX?
Thank you in advance.

[eijig]

I see 'ReActor - STATUS - Running v0.4.3-b3' in terminal and showed in Extensiton list but it dosen't show up on WebUI. Do you know a reason by any chance?

[DarkFlarePrince]

I am running stable diffusion in the cloud via RunDiffusion. I've gotten everything down except when i go to the settings, there is no models even though i already have the inswapper for the roop extension. :( do you know how to add models for this extension? i've tried /models/reactor/ and placing the inswapper there but it doesn't show up

[huhywfj]

my batch job is taking alot longer, i have a 2060 with 8 gb of ram so is there particular settings that could be slowing it down? I'm even outputting at half the size

[Sysshad]

If you already have the images saved on your harddrive then you dont need to use img->img tab. It is better to use the Extra's tab. That way you dont need to load a big SD model which steals memory for the Reactor extension. There also exists models nullmodel which when loaded uses no RAM at all.

You also can use ffmpeg from CMD/Terminal/Powershell to generate images from a movie.
create images: ffmpeg -i ".mp4" -qscale:v 2 pics/%05d.jpg
extract audio: ffmpeg -i ".mp4" -vn -acodec copy audio.aac
put it back together: ffmpeg -i pics/%05d.jpg -i audio.aac out.mp4

It is also better to use JPG (with max quality) rather then PNG, the Reactor extension has problems with PNG sometimes and crashes.

Do note that movies with decimal framerate i.e. 29.97 will be problematic to put back into a movie again (from the images), so here you first need to reencode the video into a fixed integer framerate for example 30. Handbrake is good for this.

[klpgiez]

after the installation, I can not see in the Stable diffusion category txt2img after the ControlNet the ReActor?

[rameeshariz]

is there any way i can use a mask layer to exclude some part like if a face with cooling glass

[gqbqzff]

Thank you always. I succeeded in changing my face through roop, is there a way to change my outfit and hairstyle naturally?

[moonduckmaximus]

When i drag the folder into DR from the SD batch process, the numbers come in partially out of order (1-10) clips are out of order. all of the pictures open up in the top left master area and are not one thumb nail image as shown when you do it. when i press play, its a frozen frame and wont change unless i use the slider and drag for what seems to be eternity to get the next frame. the time is way to long for the video as well. this is my first time using DR, what am i doing wrong?

[dptoproyectos]

Thanks for your detailed video. I'm trying to contact with you directly for a comissioned project. Do you have a Patreon or a way to contact you? Thanks!

[Telosx]

I'm having a problem with ReActor giving me the same exact image in img2img. I've generated a character, and I'm trying to swap my face to it, but no matter what I do with the settings, it simply does not change the output. I can do normal img2img outputs without ReActor and get changes. Any ideas?

[maximus]

Hello I am using Roop unleashed rn, its serves it purpose after some tweaking. I'd like to try this one but cannot install it, too many errors. Maybe its because I use Stability Matrix

[TheBossnia]

I don´t know if its roop or the inswapper_128.onnx but they don´t understand very well faces doing NSFW things... any of you know of a better model or application that does a better job with NSFW faces?

[Faded_Fragger]

how do I install reactor if Im running stable diffusion from a notebook on jupiterlab?

[maddydon]

Hi, i have to ask, what workload do i need to download when installing the visual studio? it seems its not working. i still cant see reActor on my webui. and i have not installed roop.

[pvqmlfz]

thanks for your help, reactor is running but I think I still miss :

onnx>=1.14.0
onnxruntime>=1.15.1
I really don't know how to install them

I get this problem where Reactor wil skip any image/frame of a video that doesn't has a face in it. So when I put it back into video(output) the audio will be out of synch and video will be all choppy. is there anyway to solve this problem?