V Express 1 Click AI Talking Avatar Generator Like D ID Massed Compute RunPod and Kaggle Guide

V-Express 1-Click AI Talking Avatar Generator - Like D-ID - Massed Compute, RunPod & Kaggle Guide

Full tutorial link > https://www.youtube.com/watch?v=GXBiqJOc9FE

V-Express tutorial for cloud (Massed Compute, RunPod & free Kaggle). Ever wished your static images could talk like magic? Meet V-Express, the groundbreaking open-source and free tool that breathes life into your photos! Whether you have an audio clip or a video, V-Express animates your images to create stunning talking avatars. Just like the acclaimed D-ID Avatar, Wav2Lip, and Avatarify, V-Express turns your still photos into dynamic, speaking personas, but with a twist—it's completely open-source and free to use! With seamless audio integration and the ability to mimic video expressions, V-Express offers an unparalleled experience without any cost or restrictions. Experience the future of digital avatars today—let's dive into how you can get started with V-Express and watch your images come alive!

1-Click V-Express Installers Scripts ⤵️

https://www.patreon.com/posts/105251204

Windows Tutorial ⤵️

https://youtu.be/xLqDTVWUSec

Hugging Face Uploading & Downloading Tutorial ⤵️

https://youtu.be/X5WVZ0NMaTg

Massed Compute Register and Login ⤵️

https://vm.massedcompute.com/signup?linkId=lp_034338&sourceId=secourses&tenantId=massed-compute

Official Rope GitHub Repository ⤵️

https://github.com/tencent-ailab/V-Express

SECourses Discord Channel to Get Full Support ⤵️

https://discord.com/servers/software-engineering-courses-secourses-772774097734074388

00:00:00 Introduction to the V-Express with demo showcase

00:02:35 How to download installer scripts

00:02:54 How to install V-Express on Massed Compute with an extremely cheap price for 48GB RTX A6000 cloud machine

00:04:25 How to install and setup ThinLinc client to control Massed Compute virtual cloud machine

00:05:05 How to setup synchronization folder on ThinLinc client to transfer files between your PC and remote cloud machine

00:06:05 How to connect Massed Compute initialized virtual machine (VM) and start installing V-Express on Massed Compute

00:06:32 Amazing special features of Massed Compute SECourses image / machine

00:07:05 How to transfer installer files to start installation on Massed Compute

00:09:05 How to start / re-start V-Express on Massed Compute after installed and use it to animate images into talking avatars

00:13:11 Where are the generated video files are saved and how to download them to our computer from remote cloud machine

00:14:11 Generating a 1-minute long animated video and how much VRAM it uses

00:15:36 How to install V-Express image to talking avatars APP on a RunPod cloud virtual machine

00:16:10 Which Pod selection settings and configuration you should do

00:16:44 Which RunPod template you have to use for V-Express APP

00:17:46 How to upload installer files to the RunPod machine and start installation process

00:18:23 How to start / re-start V-Express on RunPod after installed and use it to animate images into talking avatars

00:20:06 How to download all of the generated videos on RunPod to your computer

00:20:48 How to install and use V-Express on a Free Kaggle account with our Kaggle notebook2

00:24:18 How to download generated videos on Kaggle to your computer at once

🚨 You Won't BELIEVE What This A.I. Can Do! 🤖 Introducing the MIND-BLOWING Talking Avatar Generator That Will Leave You SPEECHLESS! 😱

Prepare to have your mind absolutely BLOWN by the most incredible A.I. technology you've ever seen! 🤯 We're talking about a revolutionary Talking Avatar Generator that creates stunningly realistic video avatars with fully synced audio - and the results are nothing short of JAW-DROPPING! 😲

But the BEST part? You can use this cutting-edge tool for SUPER CHEAP on INSANELY POWERFUL cloud GPUs like the RTX A6000 🚀 - we're talking just 31 CENTS per hour! 💸 Or if you're feeling EXTRA THRIFTY, you can even use it FOR FREE with our exclusive Kaggle Notebook! 🎉

Don't waste THOUSANDS on expensive CGI when you can create STUNNING, PROFESSIONAL-GRADE talking avatars in MINUTES with our one-click installation on Windows, Massed Compute, RunPod, and Kaggle! 💻 Our STEP-BY-STEP tutorials make it FOOLPROOF, even if you're a total beginner! 🙌

Seriously, the quality of these A.I. avatars will leave you ASTOUNDED 🤩 - it's like something straight out of a Hollywood blockbuster! 🎥 You won't be able to tell the difference between our virtual humans and REAL actors! 😎

But don't just take our word for it - we've got TONS of mind-melting demo videos and test configurations for you to try out yourself! 🎬 Trust us, once you see the UNBELIEVABLE results this Talking Avatar Generator can achieve, you'll be HOOKED! 🎣

So what are you waiting for? 🏃‍♂️💨 Smash that link in the description and prepare to have your world ROCKED by the FUTURE of A.I.! 🔮 And don't forget to obliterate that like button and SLAM subscribe for more MIND-BOGGLING A.I. content! 🔥

Video Transcription

• 00:00:00 The interesting question is, just because  something thinks differently from you,  

• 00:00:04 does that mean it's not thinking? Only audio  input and a fully talking video avatar. We  

• 00:00:11 allow for humans to have such divergences from  one another. You like strawberries, I hate ice  

• 00:00:17 skating, you cry at sad films, I am allergic to  pollen. What is the point of different tastes,  

• 00:00:25 different preferences if not to say that  our brains work differently, that we think  

• 00:00:32 differently? And you can also provide a driving  video and generate even better talking avatars.  

• 00:00:38 So today I am going to show you how to use this  amazing newest V-Express open source application  

• 00:00:46 on cloud services if you don't have a powerful  GPU. If you are following me, I already have  

• 00:00:52 published the Windows tutorial that is running  on your local computer. So if you have a powerful  

• 00:00:58 GPU, you can already use this application on  Windows with one-click installers; moreover,  

• 00:01:04 you still should watch the Windows tutorial  because I explain in more details how to use this  

• 00:01:09 application on Windows tutorial. So today I am  going to show you how to one-click install and use  

• 00:01:15 this application on Massed Compute with an amazing  price: 31 cents per hour on a very powerful GPU  

• 00:01:22 RTX A6000. You see it is running on the Massed  Compute. And it is even able to generate 1-minute  

• 00:01:29 long videos. Moreover, I am also going to show  you how to install and use on RunPod machines.  

• 00:01:36 You see I am running on three instances on RunPod  to generate multiple videos at once. Furthermore,  

• 00:01:43 if you don't want to pay to the cloud services,  I also prepared a free Kaggle account notebook.  

• 00:01:50 So I will show you how to use this application on  a free Kaggle account as well. Also, I have done  

• 00:01:56 more tests on the cloud services, and I am sharing  the test results and their configuration. So you  

• 00:02:03 can download the test file and look at each one  of the results and their configuration and decide  

• 00:02:08 which one is working best for you. Moreover,  I have prepared more demo input images and  

• 00:02:14 input videos so you can easily test the different  configuration on these files as well. And finally,  

• 00:02:22 I have updated this Patreon post, added some new  information and the new links. So everything we  

• 00:02:29 need is shared in this post. The link of this  post will be in the description of the video.  

• 00:02:35 Please read this post then go to the very bottom  of the post and you will see the installer's zip  

• 00:02:40 file which is V-Express installer's version 4.  Let's just download it. Then let's extract it  

• 00:02:46 into our downloads folder. Enter inside the  extraction folder and you will see all of  

• 00:02:53 the files here. I will begin with installing on  Massed Compute which is my newest favorite cloud  

• 00:03:00 service provider. To install on Massed Compute  just double-click the Massed Compute instructions  

• 00:03:05 read txt file. And all of the instructions are  here. First of all, please register your Massed  

• 00:03:11 Compute account or log in by double-clicking this  link. After you did that, go to the billing and  

• 00:03:17 set your credits from here. Then we will deploy  a machine. So in this interface, we are going to  

• 00:03:24 select A6000 GPU. You see RTX A6000. Why? Because  we have a special coupon for this GPU and this GPU  

• 00:03:32 is very powerful. It has 48 gigabytes VRAM. Then  from the category select creator and from image  

• 00:03:37 select SECourses. This is our virtual machine.  This is mandatory to select and to get half price  

• 00:03:43 you see currently it is 62 cents per hour. Type  here SECourses like this and click verify and  

• 00:03:49 you see now the price is 31 cents per hour. This  deal is amazing for this GPU. Moreover, we get 48  

• 00:03:57 gigabytes RAM. And 256 gigabytes hard drive. For  this machine, click deploy. Once you click deploy,  

• 00:04:05 just wait for the machine to be initialized. You  see this is the new one that we have initialized.  

• 00:04:10 I also have another running instance right now.  And let's delete the previously running instance.  

• 00:04:16 You see there is a delete button. Once you delete  the instance, all of the data will be lost. Pay  

• 00:04:20 attention to that. Let's terminate. And meanwhile,  this virtual machine is getting initialized. Let's  

• 00:04:26 also install the ThinLinc client if you haven't  yet. You see there is a ThinLinc client here  

• 00:04:31 link. I click it this arrow to open here. And when  you go to the ThinLinc client downloads, download  

• 00:04:37 according to your computer. If you are using Mac,  download the Mac version. If you are using Linux,  

• 00:04:43 download the Linux version. I am on Windows, so I  am going to download Windows. Open the downloaded  

• 00:04:49 .exe file. It will ask you permission. Click yes.  Installation is so easy. Just click yes. Next,  

• 00:04:56 install. And then run ThinLinc client. Finish.  So we are ready to use. However, before using it,  

• 00:05:02 set up your synchronization driver. You see  there are options and advanced. If you don't  

• 00:05:08 see options, click the advanced button here.  Click options. Go to the local devices. Uncheck  

• 00:05:15 all of the checkboxes and just check the drives.  And click details. You should add a new folder  

• 00:05:21 for synchronization. I am going to add my folder.  So I will remove and add again. Click this three  

• 00:05:28 dots icon. Select the folder where you want files  to be synchronized with the cloud machine or with  

• 00:05:36 your computer. So I am going to select this Mass  Compute folder on my hard drive. Click OK. If you  

• 00:05:43 want it read-only, you can make it read-only. So  you will be able to read these files, your local  

• 00:05:48 files from the cloud machine. But if you also  want to synchronize, I suggest you read and write  

• 00:05:55 and click OK and click. OK. Then just wait for  initialization to be completed. And the machine  

• 00:06:01 is initialized within like two minutes, maybe  less. Then copy this login IP. Go back to ThinLinc  

• 00:06:07 client and paste it. You see I just copy-paste the  IP. Then the username is Ubuntu and Ubuntu these  

• 00:06:13 usernames have to match. Then click this copy. It  will copy the password and paste it and connect.  

• 00:06:19 Then it will ask you this screen. Click continue.  Then you will get to this screen. Click start.  

• 00:06:26 Just wait. This part, it is very quick and we will  in a minute access the computer and yes. So our  

• 00:06:33 virtual machine has started. This virtual machine  is extremely special. It comes with Automatic1111  

• 00:06:40 web UI. It comes with SD Forge. It comes with one  trainer, Kohya. There is also one click updater  

• 00:06:47 scripts for them. Moreover, this machine comes  with Python 3.10.12 and git and CUDA 11.8. So you  

• 00:06:56 can use this virtual machine for any AI task. It  should work on this machine. So how are we going  

• 00:07:03 to install the V-Express on this machine? To  install it, copy these extracted files. Go to  

• 00:07:10 your synchronization folder, which was my Mass  Compute folder. Here. I am going to make a new  

• 00:07:17 folder here and let's say V-Express version 4 like  this. You can also extract into this folder and  

• 00:07:24 just copy paste it. Then go back to ThinLinc.com.  When you get this software updater message, just  

• 00:07:30 click cancel and where these files are copied. Go  to the home button here. Inside this home button,  

• 00:07:37 scroll down and you will see thindrives. Enter  inside it. So this is where our files will be  

• 00:07:44 synchronized. So you see this is my local folder  name. Enter inside it. Using it is very easy. Just  

• 00:07:51 after using one time, you will know how to use.  So you see this was my folder. Just drag and drop  

• 00:07:57 it for copying into desktop. You shouldn't  install into this synchronization folder.  

• 00:08:02 It will not work. So it is copied to the right  bottom of the screen. So I just drag it and drop  

• 00:08:08 it to the here so you can see. Enter inside the  folder. So this is where we are going to install:  

• 00:08:14 home > desktop > V-Express version 4. And inside  this folder, double click and open the Massed  

• 00:08:21 Compute instructions and copy this command. This  is the installation command. All of my scripts  

• 00:08:27 have the same logic. Then click this three dots  icon and click open terminal. So we have started  

• 00:08:34 the terminal inside this folder. Then copy paste  the command. So let's copy the command one more  

• 00:08:40 time. Sometimes copy paste may not work. With  Ctrl-c I copy. Then right click and paste. Okay,  

• 00:08:45 we didn't copy it accurately. Let's copy one more  time. So copy Ctrl-c. Right click and paste. And  

• 00:08:53 then it will start installation. You don't need  to do anything else. It will install everything  

• 00:08:57 all of the models and then we will run it with  one-click. Okay, so the installation has been  

• 00:09:02 completed. How are we going to start it? Let's  return back to instructions txt file and just  

• 00:09:09 copy this command with Ctrl-c. You see I write  everything for you. Then click this three dots  

• 00:09:15 icon one more time and open another terminal. You  can also use this terminal. It should work and  

• 00:09:21 paste it and hit enter. So I am showing this way.  So you can also restart it if you want to restart  

• 00:09:28 just close this terminal and open again and  execute the command and it will work. So we got  

• 00:09:33 both local URL and also Gradio live share. So you  can use both of them. If you open this link with  

• 00:09:40 right click and open link you can use it locally  or you can use it from the Gradio live as well.  

• 00:09:45 So I am going to show you how to use Gradio live.  I will copy this link and I will return back to  

• 00:09:50 my own computer and paste it and I will be using  this link from my computer. You can use it from  

• 00:09:56 your phone, mobile phone, your tablet wherever you  want and I will be able to use this application  

• 00:10:03 that is running on Massed Compute. I have prepared  two demo images and videos. You can download both  

• 00:10:10 of them and extract them to use. You can put  them into your synchronization folder. However,  

• 00:10:16 I will be using them from this Gradio. So just  click here to upload the image. So these are the  

• 00:10:22 demo images and videos part two. Select this one.  I will make a quick demo because you should watch  

• 00:10:28 the Windows tutorial of this application. In the  Windows tutorial, I have explained everything.  

• 00:10:34 You see this is the Windows tutorial. It is on  our channel. I will also put the link of this  

• 00:10:39 video into the description of the video so you  can quickly find it and learn how to use this  

• 00:10:44 application. Then let's select the video file.  You can also use audios. You don't need video.  

• 00:10:49 Both of them work. So let's select this one. So  this is a two second clip. I prepared all times  

• 00:10:56 of clips so you can use them, try them. Just click  play. So this is our driving clip and this is our  

• 00:11:03 driving image. I am going to use offset retarget.  This is the best working one. If you get too  

• 00:11:09 much problem with the mouth movements, you can  reduce the audio attention weight. For example,  

• 00:11:14 let's make this two and this will increase the  likeliness of the output image. I am still testing  

• 00:11:21 them and they will be included in the tests file  that I have shown you and auto crop is enabled so  

• 00:11:27 it will automatically crop this. The target video  width and height are set. You should watch the  

• 00:11:31 Windows tutorial. Then just click the generate  talking video. To see the entire process you  

• 00:11:36 should return back to the CMD window command line  interface where you have started the application.  

• 00:11:42 We started it on Massed Compute. You see it is  super fast to detect the faces because I have  

• 00:11:48 prepared everything accurately with the accurate  libraries. This is running on a cloud machine not  

• 00:11:53 on my machine. This machine has a huge amount  of VRAM, an amazing configuration let me show  

• 00:12:00 you. So I opened another window and nvitop is the  command. You see we have 48 GB VRAM memory here.  

• 00:12:08 We will see how much it will use for this video.  Another advantage of Massed Compute is that you  

• 00:12:14 can generate even very long duration videos  because it has 48 GB VRAM. So it is starting  

• 00:12:20 inference on our CMD window. We can watch the  status here. Okay, the inference video generation  

• 00:12:27 started so it is taking around 4.5 seconds per  IT and it will be completed in 30 iterations.  

• 00:12:35 Why? Because our step count is set as 30 so it  will take 30 iterations to be completed. As you  

• 00:12:41 increase the number of frames of your target  video or the duration of your target audio,  

• 00:12:47 it will use more VRAM memory. It is not exactly as  linear but it is getting more used. Actually after  

• 00:12:54 this video I will put a one minute video  here to show you. Okay, the result video  

• 00:12:59 is generated. Let's play. Machine is different.  Machine is different. Machine is different.  

• 00:13:04 Machine is different. Machine is different.  Machine is different. Okay, so we see the input  

• 00:13:09 and output. And where is this file saved and how  you are going to download it? Let's return back to  

• 00:13:15 the Massed Compute. On Massed Compute since this  is a desktop interface you can click open outputs  

• 00:13:21 folder and it will open the outputs folder.  So it is saved inside our installation folder  

• 00:13:26 inside outputs. So how can you download all of the  generated videos into your computer? Go back to  

• 00:13:32 the parent folder. Right click outputs and copy.  Then return back to the home. In the thindrives  

• 00:13:39 this is where our synchronization mass compute.  You can also upload your Google drive, OneDrive  

• 00:13:44 wherever you want and paste it there. After I  do that when I return back to my synchronization  

• 00:13:50 folder on my computer it will be there. Let's  return back to our synchronization folder. And you  

• 00:13:56 see output has arrived here and now this is on my  computer. So it is synchronized. Just open it and  

• 00:14:03 you will see it there. So one more thing before  I end the tutorial for the Massed Compute. Then  

• 00:14:09 we will move to the RunPod. Let's try 1-minute  length video. So this is our current VRAM usage  

• 00:14:16 only 33 megabytes. And let's go to here and just  upload a one minute video. It is also inside this  

• 00:14:23 example benedict 60 seconds. So let's upload it  into our Gradio live interface. Then hit generate  

• 00:14:31 talking video. By the way, you can also generate  higher resolution videos from here. Target video  

• 00:14:36 width and target video height. If you change them  make sure that to change the crop width and crop  

• 00:14:42 height as well if you are using auto crop image  and let's see how much VRAM we will use with 1  

• 00:14:49 minute video. So it is extracting faces in all of  the frames and you see there are so many frames  

• 00:14:55 in this video. So the inference has started. It  is using 40 GB VRAM for 1819 frames video. So it  

• 00:15:05 is a one minute video and it is using 40 GB and  it is not a problem because we are renting this  

• 00:15:11 machine with only 31 cents per hour as you are  seeing. Let's say you need even longer videos or  

• 00:15:18 bigger resolution than in the deployment menu you  can rent A100 GPU. You see it is 80 gigabyte and  

• 00:15:26 if you rent that one you will be able to generate  even longer and bigger resolution videos. So it is  

• 00:15:33 time to install on RunPod. To install on RunPod we  have RunPod instructions read txt file inside the  

• 00:15:41 downloaded zip file. So just open it. Then first  register or login your RunPod account. I have a  

• 00:15:47 link here. Please use that. So let's login.  If you don't have an account. Just register  

• 00:15:52 and in here click pods. You see I already have  two running pods which I am running the tests  

• 00:15:59 for you so you will be able to see which each  configuration makes then click deploy and in  

• 00:16:06 this screen the selections are super important.  I'm going to use community cloud because it is  

• 00:16:11 cheaper and I am going to set the extreme speed so  it will download faster. Now minimum use 3090 24  

• 00:16:20 gigabyte. You can also use RTX A6000, A40, or RTX  A5000. These are all usable graphic cards or the  

• 00:16:30 above ones from here. For this tutorial, I will  use RTX 4090, which is the fastest GPU. However,  

• 00:16:37 if you are going to generate longer videos, you  may get an out of VRAM error. Now, this template  

• 00:16:41 selection is super important. You see, after I  selected GPU, I need to select my template here.  

• 00:16:47 Click change template and type PyTorch. Now I am  going to change it. I am going to change it. After  

• 00:16:51 you typed PyTorch, you will get all of these  official templates. From these PyTorch templates,  

• 00:16:56 you need to select this PyTorch 2.1. Why? Because  this template has CUDA 11.8, which is mandatory  

• 00:17:04 to run it faster. So select PyTorch 2.1. This is  already written in the instructions, so you should  

• 00:17:11 always read my instructions for my scripts and  then click back again the edit template. Set the  

• 00:17:17 volume disk like 50 gigabytes; it is sufficient.  Also, if you want to use RunPod proxy connection,  

• 00:17:23 set a HTTP port here, 7861, and Set Overrides.  Then go to the bottom and deploy on demand.  

• 00:17:31 Let's go to my pods. So this is the new pod that  I am starting, let's say tutorial from here, and  

• 00:17:37 submit. And the pod connect button appeared; click  connect. Connect to the Jupyter lab. This screen  

• 00:17:43 may not appear or start immediately; it depends on  your pod. Now from here, click this icon. We will  

• 00:17:49 upload the downloaded zip files. So select the  extracted zip file and upload everything there.  

• 00:17:56 You see, I have uploaded everything. Double click  RunPod instructions read txt file. And this is  

• 00:18:02 the installation command. Copy it. Click this plus  icon. A new terminal here. Paste it. This is all.  

• 00:18:08 It will install everything automatically for us.  All of my scripts are working with the same logic.  

• 00:18:13 So once you learn how to run one of them, you can  run all of them. Just follow the instructions. So  

• 00:18:19 the installation has been completed. Let's  return back to the instructions file.  

• 00:18:23 Select this part. And open a new terminal. Just  copy-paste it. This will start the application.  

• 00:18:29 It will also give us a Gradio live share  because we use the share command. You see,  

• 00:18:34 we can connect from Gradio live. And also, you  can connect from the HTTP service port 7861  

• 00:18:42 because it starts on that port by default.  So you can use either way. For this one,  

• 00:18:47 I will do a quick test. So let's select  the example files. So let's pick this one,  

• 00:18:53 for example. And let's use a 7-second video file.  I am going to use offset retarget. And let's make  

• 00:19:00 the audio attention weight 2. And let's make the  reference attention weight 1. Just wait for files  

• 00:19:06 to be completely uploaded. Once it is done,  click generate talking video. I will also update  

• 00:19:13 this application and add a windows tutorial  link. And let's see the speed on 4090. So,  

• 00:19:19 it is first processing the frames. We have 233  frames for this video. So the inference has  

• 00:19:26 started. It is taking 11 seconds per it for 233  frames video. It is about a 7-second video. This  

• 00:19:36 is the speed of RTX 4090. So it is taking around  5 minutes to generate a 7-second video on RTX  

• 00:19:44 4090. It is not very fast but it is not very slow  either. So the video on RunPod has been generated.  

• 00:19:51 Let's play it. The interesting question is just  because something thinks differently from you,  

• 00:19:57 does that mean it's not thinking? Okay, it  is decent. So how you can download all of the  

• 00:20:03 generated videos on RunPod. Go to the V-Express  folder. It is inside the workspace. In here,  

• 00:20:10 you will see outputs. Right-click and download as  an archive. This is the easiest way. And you see,  

• 00:20:15 it is downloaded. Alternatively, you can use  RunPodctl to download all of them. Watch this  

• 00:20:22 amazing tutorial to learn how to use RunPodctl on  your computer. It is also an alternative way of  

• 00:20:28 downloading. One another thing is that you can use  Hugging Face uploading. In this amazing tutorial  

• 00:20:35 video, I have shown how to download and upload  files and folders to the Hugging Face. So you  

• 00:20:41 can watch this amazing tutorial to learn that as  well. Now, the final thing is that using Kaggle.  

• 00:20:48 To use Kaggle, go to Kaggle.com. This is the  slowest way, but it is completely free. Login  

• 00:20:55 with your account and activate your phone number.  Then click this create button. Click new notebook.  

• 00:21:02 In this interface, select the accelerator from  here as GPU P100. GPU P100 has more VRAM than the  

• 00:21:12 GPU T4. That is why we are using it. Then go to  the file. Click import notebook. Drag and drop or  

• 00:21:19 select the notebook file. Click browse files. Go  to the inside of your downloaded folder, you see,  

• 00:21:26 and select the free Kaggle notebook. Then once you  have selected the notebook, click import. And you  

• 00:21:31 will get to this screen. On this screen, you will  get settings updated. Click OK. It is fine. If  

• 00:21:37 you don't get it, it is also fine. Then click this  start session. Wait for the session to be started.  

• 00:21:43 If you are not able to select the accelerator,  that means your account is not phone. Then make  

• 00:21:48 sure that the internet on is selected here. Wait  for the kernel to be started. Once you get this  

• 00:21:55 green icon, click here, and you will see your  session with GPU 100 and the RAM and the memory  

• 00:22:02 used. Then click this play icon; it will install  everything automatically for you. You don't need  

• 00:22:08 to do anything else. Just wait patiently. If you  scroll down, you will see the installation logs  

• 00:22:14 like this. If it gets broken. Just message me from  Patreon, and hopefully, I will fix it immediately.  

• 00:22:20 Once the Kaggle installation has been completed,  you will see that this cancel button disappeared.  

• 00:22:26 So the installation has been completed. You see  there is no more cancel button. In the below,  

• 00:22:32 there are several steps that you have to follow to  use the web UI on Kaggle. Unfortunately, it is not  

• 00:22:39 as just Gradio share. It is blocked. So read the  instructions and apply them. And as a next step,  

• 00:22:47 we are going to execute this cell; let me show  you. So once you have applied the above steps,  

• 00:22:52 there is also a tutorial that is showing you how  to do that. A video tutorial. All you need to do  

• 00:22:58 is just click this cell. So I click this cell.  Just wait for the local URL to be started. Okay,  

• 00:23:05 running on the local URL started. Now visit the  link that you got in the above after following the  

• 00:23:11 instructions. And there you will see visit  site. After you click it, visit the site;  

• 00:23:16 you will get our interface. On Kaggle, we  have limited VRAM, so I am going to test  

• 00:23:22 with a smaller video. Let's select our image.  I am going to use this one. And I am going to  

• 00:23:28 select a smaller video, Benedict 2 seconds like  this. Let's select the offset retarget, and let's  

• 00:23:35 make both attentions 1. I think this is working  best. And click generate talking video. Then when  

• 00:23:41 you return back to your Kaggle. You will see how  it is working. It will process everything. So this  

• 00:23:47 video is 61 frames. So the inference has started,  and it is using 12.3 gigabytes of VRAM for 61  

• 00:23:55 frames video. You probably can go to 2-3 minutes,  maybe 4-5 minutes video input. But after that,  

• 00:24:03 you may get out of VRAM error. So you should  use the Massed Compute instructions, and you  

• 00:24:09 can then generate even 1-minute videos or even  longer. So our video is generated is generated on  

• 00:24:16 Kaggle as well. How you can download all of the  generated images on Kaggle. Return back to the  

• 00:24:22 Kaggle notebook, and to download all of them, we  are going to hit the cancel run button because we  

• 00:24:28 are going to execute this cell. So let's hit  the cancel run button. Then click this cell.  

• 00:24:34 This cell is going to zip all of the video files  into the working directory. So on the right panel,  

• 00:24:42 click here. And you see now we have a videos.zip  file. Whenever you click this cell, it will re-zip  

• 00:24:49 all of the videos. Then from this three dots icon  here, click and click download. And this way,  

• 00:24:56 you can zip and download all of the videos that  you have generated on a Kaggle account. And also,  

• 00:25:04 you can of course click this download icon here  and download videos directly to your computer.  

• 00:25:10 And right-click videos and you can also save  video as. And download videos and save them  

• 00:25:16 into any location you want. Thank you very much  for watching. Please subscribe to our channel.  

• 00:25:21 Leave a comment. Like. Also, when you click  here. More links. You will find our Discord  

• 00:25:28 channel. Please join our Discord channel.  You don't have to be my Patreon member.  

• 00:25:33 You see we have over seven point five thousand  members. Currently, about nine hundred members  

• 00:25:39 are online. Click join server. Ask me any  questions. You can reply to this video and  

• 00:25:45 also message me from Patreon. Hopefully,  see you in another amazing tutorial video.