IDM VTON The Most Amazing Virtual Clothing Try On Application RunPod Massed Compute Kaggle

IDM-VTON: The Most Amazing Virtual Clothing Try On Application - RunPod - Massed Compute - Kaggle

Full tutorial link > https://www.youtube.com/watch?v=LeHfgq_lAXU

In this video, I will walk through you how to install and use the amazing IDM-VTON (Virtual Try On Literally Anything) on Massed Compute, RunPod and a free Kaggle account notebook. So if you don't have a strong GPU to use IDM-VTON locally, this is the tutorial you need. We also have a local installer that works even with 8GB GPUs. This VTON is the most amazing and advanced one. It is state of the art. Moreover, Massed Compute giving us a magnificent deal. You can also follow Massed Compute installation to install and use it on your local Linux machine.

Virtual Try-on (IDM-VTON) 1 Click Installers ⤵️

https://www.patreon.com/posts/virtual-try-on-1-103022942

Register Massed Compute From Below Link (could be necessary to use our Special Coupon for A6000 GPU for 31 cents per hour) ⤵️

https://bit.ly/Furkan-Gözükara

Full Windows Install Tutorial ⤵️

https://youtu.be/m4pcIeAVQD0

00:00:00 Introduction to the IDM-VTON (Virtual Try On Anything)

00:02:12 How to download 1 click installer scripts

00:02:39 How to install and use IDM Virtual Try On APP on Massed Compute cloud computing service

00:03:56 How to download, install and setup ThinLinc client to connect your Massed Compute virtual machine

00:04:15 How to setup ThinLinc client folder synchronization to transfer folders and files between your local computer and remote Massed Compute virtual machine

00:05:00 How to connect your remote Massed Compute virtual machine via ThinLinc

00:05:19 What does End Existing Session option do

00:05:40 How to install IDM-VTON inside Massed Compute Ubuntu desktop interface

00:06:01 How to transfer files from your computer to Massed Compute cloud machine

00:06:28 How to execute install commands on Massed Compute

00:07:15 Installation on Massed Compute has been completed and how to understand and verify it

00:07:35 How to start VTON APP after installation has been completed on Massed Compute

00:08:34 How to use VTON APP on my local computer via Gradio live share that runs on cloud Massed Compute machine

00:08:52 How to upload images onto VTON APP and use the app

00:09:14 How to use Open Outputs folder and where it will work

00:10:43 Using open outputs folder inside Massed Compute VM locally

00:10:54 Where the generated images are saved by default

00:11:05 How to fix and handle Gradio related bugs and errors that you might encounter

00:11:34 Importance and effect of seed

00:11:55 How to download all of the generated images on Massed Compute

00:11:42 How to install IDM-VTON (virtual try on anything) on RunPod

00:13:00 Special configuration I make on RunPod to get very best and powerful machines

00:13:48 How to add proxy port to RunPod template to connect Gradio apps through RunPod proxy

00:14:54 How to upload installer scripts to RunPod workspace via JupyterLab and start installing

00:15:30 Installation on RunPod completed so how to start the VTON app and use it

00:16:08 How to connect VTON app through RunPod proxy since Gradio live share didn't work

00:17:04 What happens if your Gradio interface gives error while running / generating images

00:17:39 Where the generated images are saved on RunPod

00:17:50 How to download all images generated on RunPod at once

00:20:12 How to install and use VTON on a Free Kaggle account

00:22:41 How to start VTON on Kaggle with 8bit mode to make it work and start using it

00:26:06 How to download all of the generated images on Kaggle at once

In this comprehensive tutorial, I guide you through the setup and use of an innovative Virtual Try-On application across several platforms, including Massed Compute, RunPod, and Kaggle. Discover how to seamlessly transfer clothing and hairstyles to your images using this powerful tool. Learn how to leverage the cost-effective RTX A6000 GPU on Massed Compute at just 31 cents per hour, handle installations on RunPod despite some Gradio bugs, and utilize Kaggle's free notebooks for your projects. This video includes step-by-step instructions, troubleshooting tips, and links to all necessary resources and scripts. Whether you're a beginner or an experienced user, this tutorial ensures you can maximize the capabilities of the Virtual Try-On technology on any platform. Join me to enhance your digital wardrobe effortlessly!

Improving Diffusion Models for Authentic Virtual Try-on in the Wild

Paper : https://arxiv.org/pdf/2403.05139v2

Video Transcription

• 00:00:00 Greetings everyone. Today I am going to introduce  you the newest Virtual Try On application that  

• 00:00:08 has never been seen like. This application is able  to transfer literally anything with one click.  

• 00:00:16 You see, I did put the image of the Elon Musk. Let  me remove the mask that I have made and I used  

• 00:00:24 this image as a input reference to transfer  the clothing and you see it has transferred.  

• 00:00:30 It can transfer even the hair of the Brad Pitt  into my own DreamBooth generated image and you see  

• 00:00:38 the hair is transferred to me. Let me undo and  this was my original hair as you are seeing.

• 00:00:44 It can transfer the clothing of the people. You  see the input image is like this and this is the  

• 00:00:51 clothing reference, and we got the output image.  And of course, it can also transfer any clothing  

• 00:00:58 that you give like this. So today I am going to  show you how to install and use this application  

• 00:01:03 on Massed Compute which is extremely cheap,  only 31 cents per hour for RTX A6000 GPU. I'm  

• 00:01:11 also going to show that how to install and use it  on RunPod. However, RunPod is not working as good  

• 00:01:18 as Massed Compute but you will see how to install  and use it. And finally, I'm going to show you how  

• 00:01:24 you can use it on a free Kaggle notebook without  paying any money to the GPU. If you are wondering  

• 00:01:30 how you can install and use it on your computer,  we already have a one-click installer for Windows.  

• 00:01:37 You are seeing right now we already have a  tutorial for that. Actually, I suggest you watch  

• 00:01:43 that tutorial first, then watch this one. When you  go to our channel and when you go to our videos,  

• 00:01:51 you will see that this is the tutorial that I made  for Windows. So in this tutorial, I will begin with  

• 00:01:57 showing first Massed Compute, then I am going to  show you RunPod, and then Kaggle. If you look at  

• 00:02:04 the description of this video, you will find the  video chapters when they are starting, but you  

• 00:02:09 can watch the entire video if you prefer. So all  the scripts to install everything automatically  

• 00:02:17 on RunPod, Massed Compute, and use on Kaggle is  included in this attached zip file. It is at the  

• 00:02:25 bottom of the post. Go here, the link of this  post will be in the description of the video,  

• 00:02:30 then extract it anywhere that you want. I am  extracting here like this and here are all  

• 00:02:35 the installers. So how are we going to use it?  First, I will begin with Massed Compute. So open  

• 00:02:42 Massed Compute instructions.txt file. You will  get the instructions like this. First of all,  

• 00:02:47 we need to register and start our Massed Compute  instance. Currently, I am registered so I just  

• 00:02:54 logged into my account. To register, please use  this link, it is super important, and then we will  

• 00:03:00 use my coupon to get very cheap prices. On this  screen, go to the deploy menu and in here select  

• 00:03:07 A6000 GPU. So there are two A6000 GPUs. You see  alt config and A6000. Alt config has lesser RAM,  

• 00:03:17 so you will encounter problems with many of the  scripts, so I don't suggest it. Use A6000, this  

• 00:03:23 one has 48 GB GPU. Then from the category, select  creator, from the image, select SECourses and then  

• 00:03:32 use our coupon. This will make the price half.  Click verify, you will see that this amazing,  

• 00:03:39 very powerful GPU A6000 is only 31 cents per hour.  This is amazing. Then click deploy. Then you will  

• 00:03:47 get to this page and we will wait for our machine  to be started. We need to use ThinLinc client  

• 00:03:54 to connect to our remote machine. Click here if  you haven't installed it yet. Download according  

• 00:04:00 to your operating system. I am on Windows, so  download this one. It is so simple to install.  

• 00:04:05 Double-click, then click yes, next, install,  next, that's all, and then just run ThinLinc  

• 00:04:13 client. And now this screen is very important.  Click advanced to open these advanced tabs,  

• 00:04:18 click options. In here, go to the local devices.  In here you will see drives, click details, and in  

• 00:04:25 here we are going to add a folder to synchronize  between our computer and the remote machine.  

• 00:04:31 Click add. From here, it wants you to pick a  folder. I am going to pick a folder inside my R  

• 00:04:38 drive. Inside here I have a Massed Compute folder  which is here massed compute. Click okay, then make it read and  

• 00:04:46 write if you want to synchronize folders in both  ways. Click okay, click okay. Then wait for this  

• 00:04:53 initializing status to be initialized. The machine  is started. You see now the status is running.  

• 00:05:00 So copy this link like this, paste it here. Be  aware that the 300 port is not written here,  

• 00:05:07 it is just written like an IP like this. Copy  the password and paste it here. You see the  

• 00:05:13 username is ubuntu and ubuntu, and then click  connect. Then click continue. If you select  

• 00:05:19 this "End existing session", it will restart your  machine, so don't select it unless you needed it.  

• 00:05:24 It may be needed in some rare cases, like your  drive is not synchronizing. Then click start  

• 00:05:30 here. Then just wait a little bit, it is pretty  fast. Massed Compute machines are really, really  

• 00:05:36 great. Then the machine is started and we are  connected to our machine. You see it has a desktop  

• 00:05:41 interface like we are using on our computer.  So then when we go to the home and inside here,  

• 00:05:48 we will see the thindrives, and in here you  see my folder from my local computer is here.  

• 00:05:55 Double-click it and you will see the installer  files that we did put inside there. How did you  

• 00:06:01 put? So from this, copy Massed Compute install  files. Copy them, go to your drive where you  

• 00:06:09 set for synchronization. In here I have a Massed  Compute, delete them. So after I deleted them,  

• 00:06:16 you will see that they will disappear because it  will be synchronized. Then I will put them again  

• 00:06:21 like this. Put the latest version. Then when you  F5 and refresh it, it should arrive. Yes, okay.  

• 00:06:28 Then copy these two. Select, copy, right-click,  make a new folder on the desktop. So new folder,  

• 00:06:35 let's say vton_ai and click okay. When you  get this software update, just click cancel,  

• 00:06:42 you don't need it, and then double-click. This  virtual machine is amazing. We have by default  

• 00:06:48 Python 3.10, so all of the AI applications  work. Then paste the files inside this new  

• 00:06:54 folder. Double-click and open the instructions,  select the installer command like this and copy.  

• 00:07:01 Then return back to the folder and click these  three dots icon, click open terminal, paste it,  

• 00:07:07 and it will automatically install everything for  us. Just relax and wait for the installation to be  

• 00:07:13 completed. So the installation has been completed.  You see successfully installed the latest library  

• 00:07:19 and the installation is ended. You can scroll up  and check whether there are any errors or not.  

• 00:07:26 This is really, really important for you to check  if any error occurred or not, but you see there  

• 00:07:31 are no errors. It was a perfect installation. Then  return back to instructions and copy this part.  

• 00:07:38 This is where we will start the virtual try-on  application. We are going to execute this command  

• 00:07:45 inside this folder you see where we have copied the  installation files. Open a new terminal here, just  

• 00:07:52 paste the command that you have copied, and it  will start the application for us. When you first  

• 00:07:59 time run it, it will download necessary models  automatically. Then it will not download. So you  

• 00:08:06 see, on Massed Compute, you can use it either on  the local URL like this, or you can use it with  

• 00:08:11 the Gradio live share. Currently, the locally  installed browser is Firefox. Firefox is not  

• 00:08:19 the best one to use. Google Chrome is working best  with Gradio applications, so I am going to use it  

• 00:08:25 on my computer. Moreover, due to Gradio bugs,  these examples may not work, so I suggest you  

• 00:08:32 upload your own images. I am going to copy this  URL. I am returning back to my computer, paste it.  

• 00:08:39 To use this on your computer, you don't need any  hardware. This is running on Massed Compute, not  

• 00:08:45 on my computer. Then to upload your initial image,  click this icon, not here. Here is not working  

• 00:08:50 because this is a special image editor. Click this  icon, upload your initial image. Then upload the  

• 00:08:58 clothing that you want to merge. You see, it is  very robust. It even works with these kinds of  

• 00:09:04 images. Describe it, full plate armor, and click  try on. Now some people are getting confused with  

• 00:09:12 "Open outputs folder". The "Open outputs folder"  will work only on desktop operating systems,  

• 00:09:19 so it will not work when you are using  Gradio live share on your personal computer,  

• 00:09:24 or when you are using the our application on RunPod or  on Kaggle. If you are using it on your computer,  

• 00:09:31 it will work. Moreover, it should work when we  are running it on the local URL inside the Massed  

• 00:09:39 Compute. I will show in a moment once this image  is generated. Since we are running it first time,  

• 00:09:45 it is downloading the necessary models as you  are seeing with a great speed, 100 megabytes  

• 00:09:51 per second. So it is almost 1 gigabit per  second. After models have been downloaded,  

• 00:09:56 we got the output image. So you see, since I have  selected auto-generated mask and auto cropping,  

• 00:10:02 it cropped and generated this mask, and according  to this mask, this is the output I have got. I  

• 00:10:09 can mask the entire image and generate again. I  have a long tutorial regarding this that I have  

• 00:10:16 shown when I am making the tutorial for Windows,  so watch this tutorial to see how to use this  

• 00:10:23 application with more features. However, it is an  amazing application. It is very robust and you can  

• 00:10:30 even change the hair, you can even change other  features. So it is up to your imagination how to  

• 00:10:37 use it. So when I click this button, it won't work  because this is not working on my computer. If I  

• 00:10:43 return back to here, refresh this page and click  this "Open outputs folder", you see the outputs  

• 00:10:51 folder is opened inside Massed Compute. So this  is where, by default, all the images are saved.  

• 00:10:59 Due to the nature of Gradio, if you encounter any  errors, refresh the page and try again. Moreover,  

• 00:11:06 when you click this "Try on" button, it will not  regenerate the image because of a bug of Gradio.  

• 00:11:13 I have reported that and hopefully, it will be  fixed. So how you are going to solve this issue?  

• 00:11:18 Click this icon, put a dot here, then click this  undo icon and click "Try on", and it will start  

• 00:11:24 generating again. You see, I had to click double  times because this is also a bug of Gradio, but it  

• 00:11:30 is working and hopefully, all of these errors will  be fixed. So I have generated 10 images by using  

• 00:11:38 "Number of images to generate" and you will see  that each image is widely different because of the  

• 00:11:45 seed. The seed makes a huge difference. And how  can you download them to your computer? You can  

• 00:11:51 either download one by one by clicking here, or a  better way is to go to the outputs folder here. So  

• 00:11:59 I will just copy the outputs folder and I will  return back to the Massed Compute folder where  

• 00:12:06 I am synchronizing with my computer. So it is  inside home, click home, click thin drives,  

• 00:12:13 and enter inside Massed Compute. This is my folder  name, you can give any name. Paste it there. This  

• 00:12:18 is inside the remote cloud machine. Then when I  return back to my own computer, you will see the  

• 00:12:26 outputs folder gets synchronized like this. So all  the images I have generated on Massed Compute will  

• 00:12:33 be synchronized to my local computer. And now it  is time to install the virtual try-on on RunPod.  

• 00:12:42 So for RunPod, open the RunPod instructions.txt  file. Please use my register link, I appreciate  

• 00:12:49 that, so that I can get more credits to do more  tutorials for you. Log in to your account after  

• 00:12:55 registered, enter your credits, and then go to the  pods, click deploy. From here, select community  

• 00:13:02 cloud or secure cloud according to your needs.  I am using community cloud, much better for me,  

• 00:13:08 extreme speed. And now I will show you very  crucial settings. I select NVME for faster disk  

• 00:13:16 and I select 48 gigabytes to get a good machine.  So these are the settings that I use. Then select  

• 00:13:23 your GPU according to your speed need. 4090 is  extremely fast, or you can use 3090 as well,  

• 00:13:31 depending on your budget. So I will go with 4090.  And from here, select your template type, PyTorch,  

• 00:13:38 like this, and you will get these templates.  Select 2.1, it doesn't matter. Then edit your  

• 00:13:44 template and set your volume disk. Let's set 50  gigabytes. And you can also add proxy connection,  

• 00:13:51 so the proxy connection port will be 7861, like  this. Set overrides and click deploy on demand.  

• 00:13:58 Then go to "My Pods", wait for the connect button  to appear here. I have amazing tutorials for  

• 00:14:04 RunPod on my channel. Let me show you some of  them. So on our channel, click this search icon  

• 00:14:10 and type RunPod, and you will get all the amazing  RunPod tutorials that I have. And by watching  

• 00:14:16 these tutorials, you can learn how to use RunPod  very well. Moreover, I have an excellent tutorial  

• 00:14:22 where I have explained how to use Massed Compute  in more detail. It is inside this one, "Trainer  

• 00:14:28 Tutorial". It is over two hours. You can also  watch the relevant parts of this tutorial. In my  

• 00:14:33 old tutorials, I have extremely well and detailed  video sections, video chapters, like this. So  

• 00:14:40 you can read my descriptions to find wherever you  want to watch and learn. So the connect button of  

• 00:14:47 RunPod appeared. Click connect, go to the Jupyter  Lab. Once the Jupyter Lab interface is loaded,  

• 00:14:53 we are going to upload files. Okay, to upload  files, you can use this icon, select the folder  

• 00:14:58 where you have extracted files. I'm going to  upload everything because it is just easier,  

• 00:15:02 but you can also upload only RunPod files. Click  RunPod instructions.txt, copy this with Ctrl+C,  

• 00:15:09 plus icon, terminal, paste the command, hit  enter, and just wait. And let's see the speed  

• 00:15:15 of this machine. Because of the settings that I  have made, I got a very fast machine. You see,  

• 00:15:22 4 gigabits download, 4 gigabits upload, amazing  speeds machine, this one. So the installation  

• 00:15:27 and download should be pretty fast. Just wait. So  the installation of RunPod is also completed. Then  

• 00:15:34 return back to the instructions file, copy this  link, open a new terminal like this, paste it,  

• 00:15:40 hit enter, and it should automatically start the  application. First time when we generate an image,  

• 00:15:48 it will download necessary models onto RunPod.  Moreover, if you are using an Ubuntu computer,  

• 00:15:55 you should use the automatic installers for  Massed Compute. It is more compatible with Linux  

• 00:16:02 machines than the RunPod installers. Currently  we are waiting for it to generate gradio share;  

• 00:16:08 however, it is not an issue. We can go back  to my pods, click connect, and connect to  

• 00:16:12 the http service port 7861. You see, because our  application started by default on 7860 port and,  

• 00:16:23 moreover, you see that the gradio share failed  because of the Gradio server problem. However,  

• 00:16:31 we should be able to connect from the proxy,  and yes, we connected from the proxy of the  

• 00:16:36 RunPod. So, let's just upload our example image  like this, then just upload clothing like this,  

• 00:16:43 and let's say a full-scale armor. Let's generate  10 images with random seed. Seed makes a huge  

• 00:16:51 difference. Click try on, and since this is the  first time, it will download all the models,  

• 00:16:56 and you see this machine has a huge speed for  download. It is downloading with like over one  

• 00:17:02 gigabit per second. Meanwhile, still downloading  models, the gradio interface gave an error;  

• 00:17:10 however, this is not an issue because it will  continue to download models and generate images,  

• 00:17:16 and the images will be saved inside the  outputs folder. But this is an issue when  

• 00:17:21 you use RunPod. RunPod recently has been having  huge issues with the Gradio. I reported them these  

• 00:17:28 issues multiple times, but they still  didn't fix it. So, unlike on Massed Compute,  

• 00:17:33 you will have a lot of Gradio-related issues when  you are using RunPod, unfortunately. So, all the  

• 00:17:38 images are generated. They are saved inside  the outputs folder. Let's see them. You see,  

• 00:17:44 I can double-click and open each one of them as I  want. Moreover, how can I download them? You can  

• 00:17:49 select all of them, right-click and download. You  see, it is asking me to allow multiple downloads.  

• 00:17:55 Allow them. Alternatively, a better way is you  can right-click the outputs and download as an  

• 00:18:01 archive, or you can download them with runpodctl.  If you don't know how to use runpodctl, I  

• 00:18:07 have a tutorial for that. You see, runpodctl send  outputs like this. It will give me this URL. Copy  

• 00:18:14 it, then go wherever you want to download. Let's  download inside this folder. For this to work,  

• 00:18:19 you need to have runpodctl installed on your  computer, and it will download all the images  

• 00:18:26 very fast like this. Then we can return back to  our Gradio. Currently, there is a bug in gradio,  

• 00:18:32 so when I click try on, it will not work.  However, I will put a change here and undo,  

• 00:18:37 then click try on again, and this time it will  start generating the image again. So this is a  

• 00:18:44 bug of gradio. I have reported that, and  hopefully, it will be fixed soon, and we  

• 00:18:48 should get the image here this time. So let's see  the process here. For some reason, yeah, yeah,  

• 00:18:55 currently our Gradio is bugged. Okay, it started.  I don't know why, but it takes some time. Okay,  

• 00:19:01 the speed of RTX 4090 is amazing. It only takes  about five seconds to generate one image, but  

• 00:19:09 currently, this Gradio is not working accurately.  This is why RunPod is problematic recently. The  

• 00:19:15 image is generated, but we don't see the image  here, and they still didn't fix their RunPod proxy  

• 00:19:22 system. So we don't get the image, unfortunately.  Let's refresh this screen again to see. Okay,  

• 00:19:28 let's try again quickly. Let's upload our images  and let's just type full body armor and try again.  

• 00:19:36 So this is totally related to the Gradio and the  RunPod proxy. Yes, it is taking a huge time for  

• 00:19:43 gradio to connect to the server, send the task,  and get back. It is working really slow. Moreover,  

• 00:19:51 when you get a Gradio live link on RunPod, it is  working better than the proxy because of the proxy  

• 00:19:57 system of RunPod. Yes, this time we are getting  5.3 it per second. It takes only five seconds,  

• 00:20:03 and we got the image on RunPod, and when you click  open outputs folder here, it won't work. Okay,  

• 00:20:09 now it is time to start using Kaggle. For Kaggle,  we got the Kaggle virtual try-on notebook. Go to  

• 00:20:17 Kaggle.com, then log in and register, and  verify your phone number to be able to use  

• 00:20:24 the GPUs on Kaggle. Then click create, and in  this screen, we are going to first select the  

• 00:20:31 accelerator. This is super important. Select GPU  T4 x2. This is a faster GPU. Then make sure that  

• 00:20:38 internet is on here, then click file, click import  notebook, click browse files, go to the downloads,  

• 00:20:46 and select the Kaggle notebook file like  this. I see that I have a naming error,  

• 00:20:51 but it is not important. So you will get to this  screen. So how to make it work? It is so easy.  

• 00:20:58 First of all, we need to start our session.  Wait that this appears. Once it has appeared,  

• 00:21:04 click start session. You see, it was invisible,  so I just zoomed it out so that it is now visible.  

• 00:21:11 You really need to read the descriptions that I have write on my Kaggle notebooks so that you will be able to  

• 00:21:18 run them very well. And you see, I have 30 hours  of free GPU usage for this week, and it is weekly,  

• 00:21:25 so you can use it 120 hours per month. It is  a huge time to use. So click this play icon.  

• 00:21:32 It will install everything automatically for us.  Just wait for this cell execution to be stopped.  

• 00:21:40 Once this cell execution is stopped, it means  that installation has been completed, and you  

• 00:21:44 can move forward. So, the installation has been  completed. You see there is not any more cancel  

• 00:21:51 button, and when you scroll down, you will see the  installation progress like this. You will also get  

• 00:21:58 these pip installation confliction error. It is  not important, and at the very bottom, you will  

• 00:22:05 see that the installation has been completed.  Let me show you in a second. You will also get  

• 00:22:11 this dependency resolver. It is not important, and  here, so you see it was completed. So this part,  

• 00:22:19 you need to get a token and do something special  to use the Gradio interface on Kaggle. Kaggle is  

• 00:22:26 still not allowing us to use Gradio directly;  therefore, I am not going to show this section  

• 00:22:32 on the video, but necessary instructions and an  unlisted video are here. Just read this part and  

• 00:22:41 watch that video if you can't make it work.  Once you follow the instructions and got your  

• 00:22:47 link above, you just need to execute this cell to  start the virtual try-on. You see, we are using  

• 00:22:54 8-bit mode, which has arrived recently. So with  this 8-bit mode, we are able to use the virtual  

• 00:23:00 try-on on the free Kaggle notebook. Otherwise, we  weren't able to use it. So this is our work, and  

• 00:23:07 this is a huge work. If you use the default Gradio  demo that they provide with the default libraries,  

• 00:23:14 you won't be able to run this on Kaggle. And the  application started on the local URL. However,  

• 00:23:19 this will not work. You need to click the  above link that you have got by following the  

• 00:23:26 instructions above. Now, I am going to connect to  it. So once you opened that link, click visit site,  

• 00:23:32 and you will get this virtual try-on interface.  The usage is the same on Kaggle, on RunPod,  

• 00:23:39 on Massed Compute, on Windows. We just upload  our image. Let's upload our image like this. This  

• 00:23:47 time, let's upload something else. Let's try  this clothing. Let's say a t-shirt like this,  

• 00:23:53 and let's generate five images and click try on.  Same as the other platforms, when you first time  

• 00:23:59 generate an image, it will download all the models  automatically for you. You see, the download speed  

• 00:24:05 of Kaggle is the best among all others, 250  megabytes per second. It is over two gigabits  

• 00:24:12 per second. So the models have been downloaded,  and it started generating the image. However,  

• 00:24:18 Kaggle will be the slowest one among others. Each  image generation will take a huge time compared  

• 00:24:25 to RunPod or Massed Compute because of the free  GPUs that we get on Kaggle. So you see, on Kaggle,  

• 00:24:33 it is taking 40 seconds for one image. With RunPod  RTX 4090, it takes five seconds for one image,  

• 00:24:41 and on Massed Compute, it takes only 10 seconds  for one image with the very cheap GPU. So this is  

• 00:24:49 the difference between all platforms. If you use  the free one, you get slower generation speed.  

• 00:24:56 That is how it works. If you ask my opinion,  the best one is using Massed Compute. However,  

• 00:25:03 you can also use RunPod as well, or you can use  Kaggle as well. You see, it is generating right  

• 00:25:09 now. It is taking about 40 seconds, 42 seconds  for one image, but you see the gradio is working  

• 00:25:16 much better on Kaggle because RunPod is having so  many issues with gradio. Also, on Massed Compute,  

• 00:25:23 gradio is working great, no issues. So we just  need to wait. I am now generating five images to  

• 00:25:29 see them. So the images have been generated. We  can see them like this. Of course, this t-shirt  

• 00:25:35 is hard to make with this pose because it is  not compatible, as you are seeing. So you need  

• 00:25:43 to use another image if you want to get, let's  try, for example, this one. This will be much  

• 00:25:49 more compatible. This is how you can use it with  the virtual try-on. You need to have a compatible  

• 00:25:56 clothing image and also input image. However, it  is working perfectly fine. The rest is up to you  

• 00:26:03 to figure out the best settings, parameters,  and best way to use. So how you can download  

• 00:26:09 the generated images on Kaggle? To download them,  you can use several options. First of all, you can  

• 00:26:15 click this icon to expand these folders like this,  and you will get the outputs folder here. Click to  

• 00:26:24 expand it, and you can download every image one  by one. You see, download, download, like this,  

• 00:26:31 and then you can use the output here on gradio  to download them. Let's wait for the image to  

• 00:26:36 be completed. Okay, we got the output. It is much  better than the previous one. So just click here  

• 00:26:42 to download the image one by one. Alternatively,  you can download all of them. To download all of  

• 00:26:48 them, click cancel run because you cannot execute  two cells at the same time. Then click here,  

• 00:26:55 this icon. This will zip all of the images for  you in to the. The output will be inside the Kaggle working,  

• 00:27:03 so let's refresh this folder, and yes, you see  images.zip file arrived. Just click here and  

• 00:27:09 download, and you will be able to download all of  the images. I hope you have enjoyed. Please like,  

• 00:27:16 subscribe, share, and message me from Patreon  or reply to this video, and hopefully,  

• 00:27:22 I will reply to you as soon as possible. If  any of the scripts get broken, just message  

• 00:27:28 me from Patreon, and hopefully, I will fix  it as soon as possible. Thank you so much.