Donate me at Kofi: ko-fi.com/tongtong Paypal: tongtong83@gmail.com Related Video: th-cam.com/video/0p-mCnhTDNA/w-d-xo.html My PC's Spec Intel Core i5 10400F AMD Radeon RX 6600 XT 16GB RAM Video: A Young Kid Playing The Ukelele by cottonbro studio www.pexels.com/video/a-young-kid-playing-the-ukelele-3682076/ Chapters 0:00 - Intro 0:05 - Update Roop-Unleashed 0:27 - Link to Roop-Unleashed If You Haven't Install It 0:44 - Setup the DML in Provider 1:04 - Preparing Photos For Training 1:46 - Faceset Training 2:36 - Face Swap Using Trained Faceset Roop Unleashed has just been updated to version 4.1.1. If you already installed it previously, you can update it by going to the roop-unleashed folder, typing CMD in the address bar, and pressing Enter. From there, type git pull, press Enter again, and you are good to go. To run Roop Unleashed, double-click on the `run.bat` file. However, if you haven't installed it, you can watch this video on how to install it. The link will be at the top right of the video or in the description and comments below. Please note that this tutorial is specifically made for AMD GPUs only; other GPUs like Nvidia will not work. Right now, I'm going to show you how to train a faceset for future use in Roop Unleashed. Before we begin, go to the Settings, scroll down to Provider, and choose dml from the drop-down list. Click Apply Settings and Restart Server. This ensures that it will use DirectML throughout the operation. For this tutorial, I will be using Bruce Lee's face as a demo. This is strictly for educational purposes and should not be used for any illegal operations. As you can see, I have prepared 21 photos of Bruce Lee to train his faceset. There is no rule on how many faces are needed, but I personally suggest that you prepare at least 10 photos. Make sure there are no other foreign faces in the photos, as this will affect the final outcome. In this photo, you can see two blurred faces in the background; while these won't be detected, it is still advisable to crop the photo just to be safe. Go to the “Face Management” at the top of the web UI. Scroll down until you see "Input Files." Select all the photos prepared earlier and drag them into it. After a while, all the faces that have been detected will be shown as thumbnails at the bottom. Go through each face and make sure there are no foreign faces mixed in. You can always select the photo and click "Remove selected" to remove the unwanted face. Once you are satisfied, you can click "Create/Update Faceset file." A faceset file will be created on the left side of the screen. You can download the file and rename it to something easy to remember. I saved mine as Bruce Lee. Drag the Bruce Lee faceset to the Source Files and the video you want to face swap to the Target Files. I chose Codeformer from the drop-down list in "Select post-processing." You may choose anything you prefer. The newest version has a new option called "Subsample upscale to." Higher numbers will give a clearer face. I chose 256 pixels for this tutorial. Finally, scroll all the way down and click "Start" to begin. As you can see, the conversion utilizes around 20% of CPU, 10 GB of RAM, and 60% of GPU. I'm showing this only for your reference as I have a few viewers spreading hate and scolding me because they claimed that it fully utilizes their CPU. Just saying, cursing me won't help with anything. Everyone has different PC specs and all of them work differently. It took me around 7 minutes and 50 seconds to convert a 21-second video. The original video size is 14.2 MB and after the conversion, the file size has become 36.7 MB. You can always find your converted file in the roop-unleashed, output folder. Please note that if your video has a lot of weird angles, the faceswapped result may not be perfect. I hope you enjoy the video and please consider donating to me so I can make more useful videos for you in the future.
Hi, when I try to run the program I get this error. RuntimeError: Found no NVIDIA driver on your system... How can i fix it? I have a AMD GPU and i wanted to try it on CPU
Hi My memory usage is very low only using 2 gb and generating 2 frames per second. I have more vram and ram . How to utilize them. I tried using codeformer only. Also in command prompt the execution thread is showing as 1 only. Mine have 16 threads I tried increasing it in the settings but not worked
Hey bro , I can't use DML due to low VRAM but I can use CPU in roop unleashed, is that a problem on the quality of output or it's just that it takes more time?
probably will take longer with cpu, I don't think the quality is affected since it used the same process. You can see the results if it is satisfactory
Hey bro, I am using faceset of like 10 photos but in some parts of videos, the face like blurs which make it somewhat clear that is generated by Ai, any solutions to that, btw I am using max front photos in the faceset and 4 photos from a little here and there angle
That you need after effects to edit it. All the faces will become clear after swap and post process with codeformer. Another solution is swap the same video without using codeformer and you cut and paste the scene in video editor
@@K-LITOO fatal: not a git repository (or any of the parent directories): .git is a common error when git clone is not being used in the first place so it isn't detected as a git repository. Are you sure you are not downloading the zip files in the releases section and install from there? If so, git pull will never work so you have to reinstall again every time there is an update
Donate me at
Kofi: ko-fi.com/tongtong
Paypal: tongtong83@gmail.com
Related Video: th-cam.com/video/0p-mCnhTDNA/w-d-xo.html
My PC's Spec
Intel Core i5 10400F
AMD Radeon RX 6600 XT
16GB RAM
Video:
A Young Kid Playing The Ukelele by cottonbro studio
www.pexels.com/video/a-young-kid-playing-the-ukelele-3682076/
Chapters
0:00 - Intro
0:05 - Update Roop-Unleashed
0:27 - Link to Roop-Unleashed If You Haven't Install It
0:44 - Setup the DML in Provider
1:04 - Preparing Photos For Training
1:46 - Faceset Training
2:36 - Face Swap Using Trained Faceset
Roop Unleashed has just been updated to version 4.1.1. If you already installed it previously, you can update it by going to the roop-unleashed folder, typing CMD in the address bar, and pressing Enter. From there, type git pull, press Enter again, and you are good to go. To run Roop Unleashed, double-click on the `run.bat` file. However, if you haven't installed it, you can watch this video on how to install it. The link will be at the top right of the video or in the description and comments below. Please note that this tutorial is specifically made for AMD GPUs only; other GPUs like Nvidia will not work.
Right now, I'm going to show you how to train a faceset for future use in Roop Unleashed. Before we begin, go to the Settings, scroll down to Provider, and choose dml from the drop-down list. Click Apply Settings and Restart Server. This ensures that it will use DirectML throughout the operation.
For this tutorial, I will be using Bruce Lee's face as a demo. This is strictly for educational purposes and should not be used for any illegal operations. As you can see, I have prepared 21 photos of Bruce Lee to train his faceset. There is no rule on how many faces are needed, but I personally suggest that you prepare at least 10 photos. Make sure there are no other foreign faces in the photos, as this will affect the final outcome. In this photo, you can see two blurred faces in the background; while these won't be detected, it is still advisable to crop the photo just to be safe.
Go to the “Face Management” at the top of the web UI. Scroll down until you see "Input Files." Select all the photos prepared earlier and drag them into it. After a while, all the faces that have been detected will be shown as thumbnails at the bottom. Go through each face and make sure there are no foreign faces mixed in. You can always select the photo and click "Remove selected" to remove the unwanted face. Once you are satisfied, you can click "Create/Update Faceset file." A faceset file will be created on the left side of the screen. You can download the file and rename it to something easy to remember. I saved mine as Bruce Lee.
Drag the Bruce Lee faceset to the Source Files and the video you want to face swap to the Target Files. I chose Codeformer from the drop-down list in "Select post-processing." You may choose anything you prefer. The newest version has a new option called "Subsample upscale to." Higher numbers will give a clearer face. I chose 256 pixels for this tutorial. Finally, scroll all the way down and click "Start" to begin.
As you can see, the conversion utilizes around 20% of CPU, 10 GB of RAM, and 60% of GPU. I'm showing this only for your reference as I have a few viewers spreading hate and scolding me because they claimed that it fully utilizes their CPU. Just saying, cursing me won't help with anything. Everyone has different PC specs and all of them work differently. It took me around 7 minutes and 50 seconds to convert a 21-second video. The original video size is 14.2 MB and after the conversion, the file size has become 36.7 MB.
You can always find your converted file in the roop-unleashed, output folder. Please note that if your video has a lot of weird angles, the faceswapped result may not be perfect. I hope you enjoy the video and please consider donating to me so I can make more useful videos for you in the future.
my laptop takes 1 hours for 20 seconds length, RAM 16GB and VRam 1GB, is that normal? or you have spesific settings for make it faster?
Greetings, could you also make an example with an Nvidia card and if it were possible for a low-resource PC?
I wish I could but I don't have a nvidia GPU to test and make the tutorial for it 🥹
Hi, when I try to run the program I get this error. RuntimeError: Found no NVIDIA driver on your system... How can i fix it? I have a AMD GPU and i wanted to try it on CPU
Watch this video to install roop unleashed
th-cam.com/video/0p-mCnhTDNA/w-d-xo.html
I suspect you install it wrongly to NVidia version
Hi My memory usage is very low only using 2 gb and generating 2 frames per second.
I have more vram and ram .
How to utilize them.
I tried using codeformer only.
Also in command prompt the execution thread is showing as 1 only.
Mine have 16 threads I tried increasing it in the settings but not worked
That's normal, mine only used 1.3 frames/s
does it mean it's better to use a faceset than just 1 image?.....
If 1 image doesn't give you a good result, then yes
Hey bro , I can't use DML due to low VRAM but I can use CPU in roop unleashed, is that a problem on the quality of output or it's just that it takes more time?
probably will take longer with cpu, I don't think the quality is affected since it used the same process. You can see the results if it is satisfactory
Whats the difference using faceset vs 1 picture?
More accurate face swap
Hey bro, I am using faceset of like 10 photos but in some parts of videos, the face like blurs which make it somewhat clear that is generated by Ai, any solutions to that, btw I am using max front photos in the faceset and 4 photos from a little here and there angle
That you need after effects to edit it. All the faces will become clear after swap and post process with codeformer. Another solution is swap the same video without using codeformer and you cut and paste the scene in video editor
@@tongtongrizhi I see ,but when I use faceset with a codeformer or any other post processing it throws me an error, btw am at CPU
@@snappycable4995 does all select post-processing failed?
@@tongtongrizhi Yes, I have tried them once
@@tongtongrizhi I use "none"
C:
oop-unleashed-4.0.0>git pull
fatal: not a git repository (or any of the parent directories): .git
Where you get the roop-unleashed previously?
@@tongtongrizhith-cam.com/video/0p-mCnhTDNA/w-d-xo.html
@@tongtongrizhi from your previous video where you explain how to install it only for AMD GPU
@@K-LITOO fatal: not a git repository (or any of the parent directories): .git
is a common error when git clone is not being used in the first place so it isn't detected as a git repository. Are you sure you are not downloading the zip files in the releases section and install from there? If so, git pull will never work so you have to reinstall again every time there is an update