REACTOR - Face Swap for A1111 - Full Guide! Better than Roop!
ฝัง
- เผยแพร่เมื่อ 18 พ.ค. 2024
- Reacotr Face Swap - Full A1111 Guide. Install, txt2img, img2img and inpainting guide for deep fake face swapping. This Guide shows you how to use Reactor and explains some error sources for the install.
#### Download GRADIENT WORKFLOWS here ####
/ oliviotutorials
www.youtube.com/@OlivioSarika...
#### Links from the Video ####
Gradient Control: • Gradient AI Control - ...
Perfect Face: • PERFECT FACE - NO Lora...
github.com/Gourieff/sd-webui-...
github.com/Gourieff/comfyui-r...
visualstudio.microsoft.com/do...
#### Join and Support me ####
Buy me a Coffee: www.buymeacoffee.com/oliviotu...
Join my Facebook Group: / theairevolution
Joint my Discord Group: / discord
AI Newsletter: oliviotutorials.podia.com/new...
Support me on Patreon: / sarikas
00:00 Intro
00:13 AI Rap
00:34 A1111 Install Guide
03:48 Reactor Face Swap Guide
09:08 Supporter Reward - แนวปฏิบัติและการใช้ชีวิต
#### Support me + get Rewards ####
www.patreon.com/oliviotutorials
www.youtube.com/@OlivioSarikas/membership
#### Links from the Video ####
Gradient Control: th-cam.com/video/ZW1xuOB3OMk/w-d-xo.html
Perfect Face: th-cam.com/video/9UOdXLYwhYY/w-d-xo.html
github.com/Gourieff/sd-webui-reactor
github.com/Gourieff/comfyui-reactor-node
visualstudio.microsoft.com/downloads/
👋
In text 2 image , please try reactor + ip adapter face together, just use the same face input for both, u can can even use that reference image you used in your example in img2img. Use realistic vision 5.1 with cfg at 3.5 and use more basic prompts . You will be blown away by the quality and authenticness of your face. It will even do your full beard in all its glory with no extra prompting!
I test this yesterday, I had good results. I tested too the facelab and was pretty good too. Is nice because of hard angles it can get.
Something to check if you're having trouble with errors when running the extension - make sure you disable roop in the extensions tab before running reactor,
I can't find Visual studio built tools 2022 in the installer
ReActor and Faceswaplab are the goats at the moment.
fun! thanks Olivio
It is great - Thank you for the video.
Thank you too!
Getting this work had to go through lot of pain! i somehow finally made it work
same. took me 1 hour to figure out how to get it to run. bit of a mess to install
I didn’t like roop first time I tested... These methods are better, thank you again!
hi Olivio.. I love you 😍 you always make very easy to understand tutorial you're the best
Been using this for a few weeks now and it's amazing how it can make your face from a single image. I've only tested with sdxl and it's flawless
I think earlier and better is to use a controlnet face swap with a comfy UI ...
ipAdaper face swap works even with videos.
It creates a 3D model of the source and target faces and then tries to match one with the other so the orientations are the same. It's pretty amazing.
I also had problems with getting reactor to work in Comfy so I reverted back to a1111. Swapping was I was doing in my project so I couldn't go ahead with learning Comfy. Hope that gets sorted.
Now we know what Olivio's mugshot might look like if he ever got in trouble with the law...🤣
ai gangsta for life 😂
Olivio, mate, awesome video, can you please do a tutorial on the video 2 video on stable diffusion? Animatediff and others , which one is better please
amazing
ur my master seriouly i learned lot with ur videos love u from india
Works like a charm.... ✔
Faceswaplab is better to be honnest :)
I found and installed the Visual Studio community but am not seeing where the build tools are for VS
edit: So for those of you confused on where to download the build tools it's not in the installer like he would have you believe, its in the webpage where you scroll down to find the build tools exe download
thank you. I confused that in my video
Thank You!!!!
Been using reactor for a month, it runs like a charm.
But once reactor is used, it basically slows down every other operations like simple txt2img, img2img.
Only way i found is to restart a1111 to fix it.
Finally I found compared video
hi olivio: another great vid. how about one on using SD extensions Nextview and Reactor for face swapping video?
thanks, rob
The minimum and maximum effect are reversed. 1 is maximum and 0 is minimum even though it says the opposite in ReActor.
so i am getting reActor status running but i still don't have anything in the insightface\models folder. It download the inswapper_128.onnx file in the main folder (insightface) but it didnt download the models. How do download the models?
Never mind, A1111 downloaded it when I used the extension during generation.
I've been using reactor occasionally for some weeks now and it's very good. Especiall the version specially made for nvidia gpus. Yesterday I rendered around 8k frames from a small clip with the face of mine and the outcome is astounding. (atleast if the original face is looking towards the camera, otherwise it can distort sometimes)
how do you get high quality on reactor? I've just installed it and the resolution of the output is lower than the original image which is 1024*1024
Ever tries the faceswap lab extension? I think it's even better. It can synthesize a face from a Hole batch pictures. I'm pretty sure you will love it.
Lotsa work involved but good yes. Im sure its animation focused only, havent used in a while.
Not a lot of people talk about face swap lab, when I next go to make a few more characters, I'm going to try and add some expressions to the check point
exactly, although with one reference picture they give the same (99.9% similar) result not very close to the original, but the batch can improve it by a lot
cant find a tut for it! i just installed
There is on youtube. I guess in Italian. I keep saying Facelab is way better than any other. Even better than Lora for people.@@ParvathyKapoor
Thank you for video! One question: why you still using xformers with torch 2.0? “-Opt-sdp-attention” is faster and use less vram than xformers on RTX card.
After 1 hour of installing and troubleshooting, it finaly works! hahahaha, but after the first try, I got an NSFW image with the head of Mona Lisa hahahaha
Can this deal with glasses and beards and different expressions like Roop? And how you you rate it versus the Roop extension repo? Thanks!
Thank you. I tried to install this several weeks ago without success. Manually adding the [insightface] folder fixed it for me. =)
awesome, glad i could help
Could you share a link for the insightface folder?
thank you a lot Olivio, we are all doing good progress ... how can we got the same pass of what you have? I want to test on google colab, I did the test, works but I think I got very basic interfase, do you have videos how to create or work with /comfyui ?
😊very cool
thank you
I'm stuck on the part where I can install Visual Studio Community 2022 but I don't see anything called Visual Studio Build tools 2022. Is it by another name instead?
Click the Available tab and you'll see it listed. You only have Community installed so the Installer is showing only that.
Don't know who else to ask, can you please make a video on Comfy UI inpainting at full resolution similar to how it works in A111, where it masks and renders a smaller section of the image, not the entire thing. Thanks man
hey, awesome video! My insightface folder doesnt appear, and you dont really explain how you fixed that issue. Automatic1111 works, and in the cmd window when I launch it it even says that Reactor is working, but when it opens I don't have the extra options when I scroll down.
Good video. Do you have an equivalent video of ComfyUI gradient @ 9:10 but for A1111 please?
same as with roop, reactor just doesn't show up after installing
Is it higher sampling steps can effect the quality of face swap on img2img
Love the rap!
Thank you 😍
The current model in Reactor can't be used for commercial work. Does anyone know if there is an alternative model to use for commercial use. Potentially at a cost?
I prefer to stay on FaceSwapLab because I completely bricked my A1111 (I had to install again the whole SD with A1111) when I tried to install Reactor one week ago. Maybe I did something wrong but I prefer to no taunt the devil.
Is Reactor really better or is it the same ?
Hi can you show us how to set up Comfyui ReActor because I am unable to get it to work even though I do see a folder showing it is uploaded but I can not see it or figure out how to fix the issue.
Can we select faces individually if we use image with multiple faces.
What should I do if, when generating images, a long fringe covering the eyes results in the forehead appearing darkened?
Awesome video as usual! Can you give youre "prompts" please?
any pure extension methods without installing distribution files on local drive?
hi i did everything right but i do not have the "enable"button in ReActor...do you have any ideas why?
PLEASE, help me to install Face Swap in PaperSpace. I don't have the machine to use A1111 locally therefore installing Visual Studio does not make any change in the PaperSpace environment.
my reactor Running v0.7.0-b7 on Device but no "enable box" & not single box but 2box
i cant get a beard to show up in the swap it shows up as like a 5 o'clock shadow. how do you get the beard to come in more?
ouh lala!!
I am trying to get reactor to show up, but it may i do not have insightface. How do i install Insightface?
Did you know if I can install 2 différent GPU for cumulating the Vram ?
RTX 3090 Ti Fe 24G, and RTX 2070 Super 8G. equal 32G Vram
I know it works for 3d rendering, but does it work for AI génerating ?
i doubt it, because as far as i know a single AI image and always only be rendered on a single GPU.
Any advice on how to troubleshoot the installation? I cant get reactor to launch in A111. When I start the server I get an error message to restart the server, so I restart it, and the same error occurs again. As a result reactor is never available.
I think FaceSwapLab may be better than reactor but not sure yet... the model building feature of faceswaplab seems to create better results
Can you make on video on version 6.1? This seems quite different I can't follow instructions with 6.1
Any method for a Macbook install of this? (ie is MS Visual Studio a pre-requisite, and will the Mac version of that work?)
Did you get working on Mac?
@@officialmultimedia8107 Yes. No need for the MS Visual Studio install. Macs already have the right bits for this to work without that. Jst follow the install instructions, and works fine..
Hello, I'm wondering is there any techniques to alter face without a source? I mean can we modify hair, eyes? Particular attributes only?
you can do that by changing the text prompt. look for online prompt lists for different hair styles for example. Also you could use controlnet. that does require a photo, but you could just take a photo of yourself and use that as a tracking in openpose without it actually taking the shape of your face
The results look the exact same as roop to me and from the other comments they use the same model underneath so they are in fact the exact same.
yes, but you need more than just the model. It's like saying you use android, so why are there different phones. because different developers still bring different ways to use the same thing. for example i can't remember that roop ever tracked the age and gender of a person in the photo
@@OlivioSarikasif roop developer was still working he would have added it. the co developer is now working on facefusion and face fusion has age gender etc built in. but it sucks in detection as roop algorithm is better
@@OlivioSarikas it does, but in the background.
It didnt install the inswapper_128.onnx model for me where do I get the models?
How well does Reactor work if you want to swap a face that's not from a front angle, but more like a face looking down or profile, but the face you're swapping in is just the front view?
i would always use a face from the front angle as an input, because the ai needs to see the full face features and dimensions. you can try it with a rotated face, but the likeness will not be as good. i would also suggest to use a image that does not have a too strong face expression, but if you intend to have a smile in it, simle in the image so that ai can see what your teeth look like
@@OlivioSarikas Thanks for your answer!
Consider a photo for a Visa to another country. This is what the ai needs to extrapolate the contour of the face to generate multiple instances of different profiles. Per Visa/Consulate request: Need a clear & colored photo that must be taken against solid white background in neutral expressions with face cantered with clear visibility and both shoulders should be covered & within the frame; Ears should be visible, no shadow in the picture, without glasses & jewelry. Teeth should not be visible & no white cloths shouldn't be wearing in the photo. Now the teeth are optional as people have different features when smiling. Meaning some people are recognized by their smile, doesn't have anything to do with their teeth in this case. If you follow this, you will always get a perfect conversion of photos.
How can I use A1111 and ComfyUI in the same portable sd folder?
unfortunately I don't get Visual studio build tools 2022. seems to be a new version without that feature. reactor doesn't work for me. Maybe that's the reason.
Seems like it ought to be possible to make a blender model that outputs openpose poses. Are you aware of anyone doing this? It outght to be a pretty easy geometry to rig.
can I combine reactor with expresion loras ??
i keep geting error: AttributeError: 'NoneType' object has no attribute 'get' in both SD and SDXL, can anyone help me on this?
1:58 Any idea why I don't have an A1111 folder? I have a stable-diffusion-webui folder. Is that the same thing?
yes it's the same thing
Sad to hear Roop is being discontinued!
I cant get it to look anything like the photo - used a high quality clear full on face pic I took, and no matter how I use it it just looks like a vague-ish version
I am having problems with objects over the face. Reactor doesn't overlap objects over or in front of the faces, it just draws the face over the objects. How to achieve this?
👍
haven't really been successful installing reactor and having it work for me, it always gives me errors after trying to generate an image, iv updated visual studio community, updated the pip p, and even tried installing and restating but no luck, one of the lines in the errors is TypeError: 'NoneType' object is not subscriptable what can i do?
I face the same issue
can we use this on google collaboration
Hello, I have a question. How can I use it functionally?
HALP!
I've tried in multiple ways to install this extension to A1, but nothing seems to work.
I have it checked within A1 extensions interface. The CMD gives no errors. But I can't see the drop down menu item in the main interface.
One thing I discovered just now is that within the Insightface folder you mention, I do not have all the models you talk about. It is empty, apart from the file "inswapper_128.onx". How do I get it to download and fill that folder with those models?
I sorted it by reinstalling insightface as per this guide:
VIII. (For Windows users) If you still cannot build Insightface for some reasons or just don't want to install Visual Studio or VS C++ Build Tools - do the following:
Close (stop) your SD WebUI Server if it's running
Download and put prebuilt Insightface package into the stable-diffusion-webui (or SD.Next) root folder (where you have "webui-user.bat" file)
From stable-diffusion-webui (or SD.Next) root folder run CMD and .\venv\Scripts\activate
Then update your PIP: python -m pip install -U pip
Then install Insightface: pip install insightface-0.7.3-cp310-cp310-win_amd64.whl
Enjoy!
was this video taken down and then re-instated?
Facefusion or Reactor? Which is better?
How do i get Visual Studio Tools 2022 there is no such option?
mine is generating a random person and then pasting my source face onto that. anyone know what im doing wrong? pls
They all use the same model (insightface 128), so how are they better?
Faceswaplab has built in upscaling.
@@GunnyNinja that upscaling is applied to whole image, not only the face, so the face remains low resolution as compare to whole body/image
@@masroorbabar3381 Not in my use.
@@masroorbabar3381I should clarify, mostly I am using Faceswaplab as a replacement for dreambooth. The face is rendered at the same resolution as the image. So, in img2img, you are correct. Still looks better than what I got from Roop GE(reactor).
Can you also do a "clothes swapping", like you want the person in the image to wear a pair of shoes or shirt you have just one image of?
don't you mean clothes removal? lol
@@chucknorris8704 this would probably be one of the easiest things for inpainting or img2img, tbh :D
@@chucknorris8704 that's just inpainting tho?
interesting idea. so far i don't know an automated way to do that
Paint the new clothes onto the person in the image, using Photoshop or similar. It can be a very rough sketch. The dimensions and colors are important. Then mask the area in in-painting and prompt for the different clothes.
It's much more better not to install VS-Studio. In installation instruction to Reactor there is another way, much more simple and in this case you save your disk space.
it’s exactly same model as roop with same results
but roop isn no longer being updated. it was discontinued 2 months ago!
So its still Roop? Using the 128x128 lowres insightface model. Ill pass.
I've got it "working" but the results are awful... I must be doing something wrong
2:45 You literally say it was not installed but give no pointers on how to install.
How did you do the voice filter in the intro? 😱
that's just Davinci Resolve Pitch Down ;)
Reactor just doesn't appear under control net even after checking for updates on the xtension and apply and restart UI ☹☹☹☹
same issue
I have tried many solutions. Nothing has worked out for me @@mariomills
I found the solution, so on the txt to IMG page, go all the way down, there's a menu and click on "reload UI". After this it worked. When I used the reload UI link from the "installed" tab, it doesn't work. Try this
Glad to hear that it worked for you. I reloaded the UI and still is not showing up. I saw files like w600k_r50.onnx missing from the original gitHub page link. I don't know why but for some reason the only file after the installation that was showing in my folder was inswapper_128.onnx nothing else. No buffalo_l folder and the rest of the modules were all missing. why? I have no idea. I had to search the web and find all the different modules and manually download them and put them in the proper place and still is not working. I am running the latest version of SD as well. @@mariomills
So can I take Midjourney images and use this to face swap?
I would think so, just save them to your local hard drive and then load them into Img2Img tab.
technically yes, it's just a matter of if the face in MJ has the correct anatomical sice and shape
he said both install methods did not create the insight face folder, so where did he get it and why did he jump to saying after it gets done installing...im lost with this guy
you make a new folder and call it "insightface"
I followed the steps, even though it shows in extensions, it's not available as a tab anywhere. No idea how to find it
I had the same problem. I had forgotten to select the C++ tools when installing Virtual Studio
any solution for MAC users?
What if there is an object in front of the face?
My insightface folder is empty, from where I can get it correctly installed?
I have same problem
FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\xx6\\Documents\\A1111 Web UI Autoinstaller\\stable-diffusion-webui\\models\\insightface\\inswapper_128.onnx'
I manually create Insightface folder in the '..webui\models' folder and copy inswapper_128.onnx file and it works. I am still missing subfolder named bufallo_l and dozen of .onnx files inside. Please someone put link on that folder to download manually
no answer ? 😞
No answers yet ;-;
I'm starting and can't make it work... It's not changing the face at all :(
I have an issue tho, everything works fine but i can't replace the face to an existing image, it simply creates a new image from scratch. Can anybody tell me how can I keep the same image as base and just swap the face in text to image? Ty!
Are you sure you enabled it? Also check your commandline window for any error messages
@@OlivioSarikas yes it's 100% enabled and working, the only thing is that in order to keep the consistency of the base image I only can use img to img or inpaint as you taught me. If i use txt to image, the only consistent thing is the replaced face, everything else changes! :D Anyways, ty so much for answering and REALLY I do appreciate your work and I learnt so much through you! ♥Thank you so much again for your work!
Why not use comfui and controlnet face swap model .... Works perfect.
because this is a video about reactor. just because there is green tea, people still like to hear about black tea
@@OlivioSarikas Ok .. understand 😅
You could say something more about Fooocus as it got a lot of updates since you last video.
About his prompt extension ( gpt2 ), impairing , outpaintig , his controlnet implementation , photorealistic picture implementation which works great , etc ...
Thanks for the video anyway .
You're a great man !
@@mirek190 good point. i will look into ti
@@OlivioSarikas thank you
why wouldn't you show us how you fixed insightface....
It's basically a roop, right ? Only works in low res ?
oh that means the same problem exist? like for videos@@TiTechno
@@TiTechno no, as in if there's object like hands overlapping then it fails to detect/clone face, then there are frames not cloned
Any users of A1111 on a M2 Book managed to get this to work?