SECourses Upscaler Pro Beating Topaz AI by Far With Specalized FlashVSR+ & SeedVR2.5 - Local Windows

[FurkanGozukara]

SECourses Upscaler Pro Beating Topaz AI by Far With Specalized FlashVSR+ & SeedVR2.5 - Local Windows

Full tutorial: https://www.youtube.com/watch?v=_WT4C78j5-c

🚀 Welcome to the Ultimate SECourses Upscaler Pro & Trellis 3D Tutorial!

Greetings everyone! Today, I am incredibly excited to showcase the massive new improvements and brand-new features we have added to the SECourses Upscaler Pro application. I have been working non-stop to bring you a studio-level AI video and image enhancement tool that completely redefines what is possible running locally on your own PC.

In this video, we dive deep into side-by-side comparisons between our custom FlashVSR+ upscaler, original viral social media videos, and Topaz AI. As you will see in our live slider comparisons, the SECourses Upscaler Pro is adding 10x more detail than Topaz AI, generating breathtaking, high-definition results while running highly optimized on GPUs with as little as 8GB of VRAM!

We also explore the immensely powerful SeedVR2 model for flawless 4x image upscaling, and I give you an exclusive sneak peek at our upcoming Trellis Image-to-3D application featuring fully automated UniRig 3D character rigging!

🔗 Important Links & Resources:

📥 Download the Latest SECourses Upscaler Pro Installer: [ https://www.patreon.com/posts/secourses-upscaler-pro-150202809 ]

📥 Download Trellis Image-to-3D App: [ https://www.patreon.com/posts/trellis2-app-147686623 ]

⚡ Key Features of SECourses Upscaler Pro Covered in This Video:

•• Unmatched AI Upscaling: Outperforms Topaz AI and other public repositories by adding massive amounts of true-to-life details to low-resolution inputs.

•• Optimized for Low VRAM: Run advanced models like FlashVSR+ and SeedVR2 on GPUs with 8GB VRAM using GGUF and FP8 models.

•• Scene-Based & Chunk-Based Processing: Automatically detects scene changes and processes videos in chunks. This maximizes quality and prevents VRAM crashes!

•• Full Resume Capability

•• Built-In Comparison Tools: Automatically generate custom, animated side-by-side comparison slider videos with custom text labels to showcase your upscales.

•• Advanced Video Output: Supports global RIFE frame interpolation, Face Restoration, H265 10-bit encoding, and custom audio codecs.

•• Easy Installation: 1-click install and update .bat files. No complicated coding required!

⏱️ Video Chapters & Timestamps:

00:00:00 Introduction to SECourses Upscaler Pro & Amazing New Features

00:00:29 Comparison: Original Video vs SECourses Pro Upscaled Video

00:00:50 Slider Tool Comparison: Original vs SECourses Upscaler

00:01:06 Comparison: Original Video vs Topaz AI Video Enhance

00:01:19 Slider Tool Comparison: Original vs Topaz AI Quality

00:01:35 Ultimate Comparison: SECourses Upscaler vs Topaz AI

00:02:07 Overview of FlashVSR+ Version and Custom Improvements

00:02:20 Live Studio Level Video Comparison Slider Tool Demonstration

00:03:02 Topaz AI vs Original Video Slider Demonstration in App

00:03:15 SECourses vs Topaz AI Massive Quality Difference Revealed

00:03:35 Running Locally vs Cloud GPUs (RunPod, MassedCompute)

00:03:56 Test Case 2: Viral AI Video Original vs SECourses Upscaled

00:04:17 How to Generate Custom Comparison Videos with Custom Titles

00:04:50 Test Case 2 Slider: Topaz AI vs SECourses FlashVSR+

00:05:25 Chunk-Based & Scene-Based Upscaling System Explained

00:05:50 How to Download, Install, and Update the Application Easily

00:06:34 SeedVR2 Model Tricks: Batch Size & Blocks to Swap Setup

00:07:00 Live Demonstration: 4x Image Upscaling with SeedVR2

00:07:36 Optimizing VRAM Usage with Max Blocks to Swap & GGUF Models

00:08:34 FlashVSR+ Video Upscaling Setup & Scene Detection Features

00:09:17 Downscale and 4x Upscale Resolution Configuration Explained

00:09:54 How to Use the Full Resume Capability for Long Video Upscales

00:10:15 Custom Target Resolution Settings & Target Pixel Dimensions

00:10:46 Frame Chunk Size Optimization for Maximum Speed & Quality

00:11:20 Enable DiT Tiling vs VAE Tiling for VRAM Management

00:11:58 Live FlashVSR+ Upscaling Process & Input Chunk Splitting

00:12:43 VRAM Usage Analysis: Upscaling to 4K on a 16GB GPU

00:13:18 Additional SeedVR2 Tricks: Sparse Ratio & Tile Size Limits

00:13:41 Canceling Upscale Jobs & Monitoring RAM/VRAM Release Live

00:13:53 App Features: Health Check, Global Settings & Light/Dark Theme

00:14:15 Queue System, Face Restoration & RIFE Frame Interpolation

00:14:50 Video Output Formats, Codecs & H265 10-bit Encoding Setup

00:15:17 Automatic Comparison Display & Run Metadata Saving Feature

00:16:03 Direct Video Compare Tool & Animated Slider Generation

00:16:34 Multi-Video Comparison Slider (Original vs Topaz vs SECourses)

00:17:32 Testing Default Settings for VRAM Memory Efficiency

00:18:15 FlashVSR Default Settings Setup & Upscale Demonstration

00:19:07 Verifying Low VRAM Usage Under 8GB for HD Video Upscaling

00:19:38 Upcoming Trellis Image to 3D Application Features Preview

00:19:51 UniRig 3D Rigging Integration for Character Skeletons

00:20:15 Compiling Custom 3D Libraries for Modern CUDA 13 GPUs

00:20:53 Trellis Output Formats: Skeleton FBX & Preview GLB Saves

00:21:14 Importing & Previewing 3D Models, Bones, and Textures

00:22:20 Future Updates: Multi-Stage Upscale Queue & SUPIR Integration

Video Transcription

• 00:00:00 Greetings everyone. Today I am going to show  you SECourses Upscaler Pro application. I have  

• 00:00:06 been working on this application non-stop since  the last time I have made a tutorial video. We  

• 00:00:12 have made so much, so amazing new improvements and  added new features to this application. So today,  

• 00:00:18 I am going to show you them. Plus, I will show  you new upcoming features to our Trellis image  

• 00:00:24 to 3D application. Let's quickly begin  with some comparison of our application.

• 00:00:29 The 1st is our SECourses Pro Upscaler  improvements. So the top 1 is the original video,  

• 00:00:37 the bottom 1 is our upscaled video. As you can  see, our application added so much amazing new  

• 00:00:44 details to the video. So let's see its slider  comparison version. So it is beginning from  

• 00:00:50 left to right. The left 1 is original video, the  right 1 is the upscaled video. As you can see,  

• 00:00:55 our upscaler is adding amazing amount of  new details to the video. This video went  

• 00:01:01 viral on social media, this is an AI video. So  let's compare it with the Topaz AI and see the  

• 00:01:07 result. So this time we are seeing original versus  Topaz AI. Topaz AI adding so little amount of new  

• 00:01:13 details compared to our SECourses Upscaler  Pro application, as you can see right now.

• 00:01:19 So let's see the slider version of the Topaz AI.  So here on the left we see the original video,  

• 00:01:24 on the right we see the Topaz AI version.  There is a too little amount of new details  

• 00:01:30 when we consider the comparison with our  upscaler. And finally, let's compare it  

• 00:01:35 with our upscaler. So SECourses versus Topaz  AI. Okay, the left 1 is our SECourses upscaler,  

• 00:01:42 the right 1 is the Topaz AI. Pay attention to the  details difference. I mean, our upscaler is like  

• 00:01:48 10 times better than the Topaz AI. It is working  as low as 8 gigabyte GPUs, maybe even lower. It  

• 00:01:55 is also really fast. Let's watch 1 more time.  You see the details, the amount of details in  

• 00:02:01 SECourses Upscaler Pro FlashVSR+ is much better.  It is many times better than the Topaz AI.

• 00:02:07 This FlashVSR+ version is made by me. It has so  much new improvements and features compared to the  

• 00:02:14 other public repositories that you are going to  find. Let's also see it from the video comparison  

• 00:02:20 slider. So I am going to add test 1, test 1  SECourses, and test 1 Topaz AI to compare them.  

• 00:02:26 This application is studio level. I am going  to show you some of the amazing features. So  

• 00:02:30 let's make full screen. So on the left we see the  original input, on the right we see the SECourses,  

• 00:02:36 as you can see right now. And you can play with  the slider, let's also sync it because sometimes  

• 00:02:42 it may get out of sync, and you can see how much  details our upscaler is adding. So you see, the  

• 00:02:48 right 1 is our upscaler, the left 1 is original  video. It has added so much amount of new details  

• 00:02:56 to the video. Let's see the Topaz AI version.  So now it is original versus Topaz AI. So let's  

• 00:03:03 see the slider. I mean, look at it, the Topaz AI  adds so little amount of details compared to our  

• 00:03:10 SECourses upscaler. Let's compare SECourses versus  Topaz AI. Now you will notice the difference much  

• 00:03:16 better. So the left 1 is our 1, the right 1 is the  Topaz AI, and as you can see, there is a massive  

• 00:03:22 amount of quality difference between the SECourses  FlashVSR+ upscaler and the Topaz AI upscaler.

• 00:03:30 Our upscaler is next level compared to  it. This upscaler is running locally on  

• 00:03:35 my Windows computer. You can also run it on  cloud GPUs, cloud services if you don't have  

• 00:03:40 a powerful GPU. You can also rent a powerful  GPU on RunPod or on MassedCompute or SimplePod  

• 00:03:46 and upscale much longer videos much faster into  much higher resolution than your own GPU. So our  

• 00:03:52 application has so much extra features.  Let's also see the test case 2. So 1st,  

• 00:03:57 let's begin with the original video versus our  upscaler in the original video. So this video  

• 00:04:05 went viral on the social media, as you are seeing  right now. Our upscaler is doing another amazing,  

• 00:04:11 great job. Let's see the slider version. So test 2  original versus slider with SECourses. By the way,  

• 00:04:18 I have used our application to generate these  comparison videos, I will show you that. So  

• 00:04:22 this is also a feature of our application that  you can generate such comparison videos with  

• 00:04:28 custom titles, even with custom frame rate or the  speed of the slider. As you can see, FlashVSR+ of  

• 00:04:36 our own version adding so much amazing details  compared to the original video, and I am going  

• 00:04:42 to also show you now the Topaz AI version. Let's  see the Topaz AI versus SECourses slider. So it is  

• 00:04:51 beginning, the left 1 is our upscaler, the right  1 is the Topaz AI. I mean, the Topaz AI is almost  

• 00:04:57 as if it is like the original video. There aren't  new details. This is generated with their latest  

• 00:05:03 version, Topaz AI latest version, default settings  and 4K upscale. This is what it could do maximum.

• 00:05:09 And our application is running locally as low  as 8 gigabyte GPUs, I will show you live. And  

• 00:05:15 as you can see, our upscaler is next level. I  mean, it is not even comparable to the Topaz  

• 00:05:20 AI at this quality. But these are not all the  features of our application. Our application  

• 00:05:26 works with chunk-based, scene-based upscale.  As you can see, it is splitting the video into  

• 00:05:33 scenes and upscaling each scene at a time. So it  improves quality, it adds full resume capability,  

• 00:05:40 it lets you view the upscale live during  the processing. So you can decide whether  

• 00:05:46 it is working good or not. So let's restart the  application and show you. The installation is  

• 00:05:51 also super easy. So you will be able to download  the latest zip file from this link, it will be in  

• 00:05:57 the description of the video. The latest installer  zip file, it is 2.1, but the zip file name may not  

• 00:06:03 match the latest version because you don't need  to update the zip file every time. Just running  

• 00:06:09 the install or update .bat file will do, and you  should follow requirements tutorial if you want to  

• 00:06:13 use Torch Compile. So installation and update  is just Windows install and update .bat file,  

• 00:06:19 then we will run the upscaler pro .bat file.  It will download all the models automatically,  

• 00:06:25 it will make all the installation automatically.  I will make a long video later about this.

• 00:06:29 So let me show you some of the tricks of how to  use application in this quick tutorial video. In  

• 00:06:34 the SeedVR2, the biggest trick is the batch  size and the blocks to swap. This model is  

• 00:06:41 amazing for upscaling images. For video upscale  with SeedVR2, you really need a bigger GPU,  

• 00:06:47 but for image upscale it is amazing. So I am going  to upscale this image into 4x. It is 1024 to 1024,  

• 00:06:54 now it will become 4096. Let's upscale live,  live demonstration. If you get out of VRAM error,  

• 00:07:01 just maximize the blocks to swap and you can also  enable Torch Compile, it will speed up. So this  

• 00:07:07 resolution is requiring right now 26 gigabytes  of GPU because this is a massive resolution,  

• 00:07:13 but it is done in 7 seconds. Like total 8 seconds,  or I think total will be like 20 seconds. Yes,  

• 00:07:21 it is done. So we upscaled this image from 1  megapixel into 16 megapixel within 30 seconds.  

• 00:07:28 And let's see the quality difference from left to  right, left to right. And let's also maximize the  

• 00:07:37 blocks to swap to see the both speed and the  VRAM usage. If you don't have a powerful GPU,  

• 00:07:43 I recommend you to maximize the blocks  to swap. You can also use the lower VRAM  

• 00:07:49 using models like 3 billion parameter. We are  also supporting the GGUF models or FP8 models,  

• 00:07:55 you can use any of them. So with the maximum block  swap, it is going to use like 12 gigabytes of VRAM  

• 00:08:02 for this resolution. I mean, you can reduce the  resolution and reduce the VRAM usage as well.  

• 00:08:06 This is the highest quality of an upscale.  It is still really fast. I mean it took,  

• 00:08:11 yes, it took lesser this time because the 1st was  initialization, and it is done. This will make you  

• 00:08:18 lose 0 amount of quality, it will be exactly  same quality. It will be automatically saved  

• 00:08:23 into the outputs folder with all the metadata and  everything. You can see everything here, upscaled  

• 00:08:29 file, pre-processed file if it is downscaled  or whatever, because it shows everything.

• 00:08:34 Let me show you a video in FlashVSR+. So with  the FlashVSR+, we have several options. Let's 1st  

• 00:08:42 select our test video. I am going to show you with  test 2 video. When you upload video, it will 1st  

• 00:08:49 analyze and scene detect it, as you can see. So it  will show you all the information both for images  

• 00:08:55 and videos. Okay, it is still updating. If it  doesn't refresh at the 1st time, you can refresh  

• 00:09:01 the page, I am improving this process as well.  Okay, it is done. So this video is 1320 to 936  

• 00:09:09 pixels, and currently we have set 4x upscale. This  model can do 2x or 4x. SeedVR2 can do any ratio.  

• 00:09:17 The 4x is the maximum quality, but 2x is also  very good. So it is going to upscale 4x. However,  

• 00:09:24 since our downscale option is enabled, it is going  to 1st downscale. So it is going to downscale into  

• 00:09:31 1480 to 340 pixels, then it will upscale 4x  into 1920 to 1360 pixels. And it is going to  

• 00:09:42 split it into 9 different scenes and upscale each  scene individually. Therefore, it will be maximum  

• 00:09:49 quality and you can upscale very long videos and  you can have full resume. For resuming, you just  

• 00:09:55 need to provide outputs folder path. For example,  let's say I want to resume the upscale 96. So I  

• 00:10:02 just copy paste it and start upscale, and it will  continue from which scene, which chunk it is left.  

• 00:10:09 So what other tricks that you can do? You can  set this to any resolution you want, then it will  

• 00:10:15 update all the values here. Pay attention to the  values here. So now it is going to only downscale  

• 00:10:21 to this pixel and it is going to upscale into  3840 to 2820 pixel. This is a massive resolution.

• 00:10:29 So these are the 2 tricks that you can do in this  top section. I don't recommend you to change model  

• 00:10:35 version, pipeline mode, or VAE model. These are  the best and it is working really low VRAM. So  

• 00:10:40 what other thing you can do? Frame chunk size.  As you process more frames at 1 time, it will be  

• 00:10:46 faster and higher quality. Since this is extremely  optimized model and since we are processing chunks  

• 00:10:53 based, I recommend you to increase this and test  whether you are getting out of VRAM error or not.  

• 00:10:58 And currently my maximum frame chunk processed at  a time will be 299. Even though the video is 1055  

• 00:11:07 frames, since we are automatically splitting  it into the scenes, the maximum 1 will be 299,  

• 00:11:13 average will be 118 frames, minimum will be the  40. Okay, then what other options there are? There  

• 00:11:21 is enable DiT tiling. You should enable this,  otherwise it will use a lot of VRAM. So unless  

• 00:11:26 you are working on cloud GPUs, you need to enable  this. If you disable this, it will be much faster,  

• 00:11:32 many times faster, and it will be a little  bit higher quality. Don't enable VAE tiling,  

• 00:11:38 I don't recommend it. It brings so little amount  of VRAM and it can cause errors and other issues.  

• 00:11:45 This model can also upscale images, but it is not  as good as SeedVR2 in my opinion, you can test it.

• 00:11:52 So now when I click start upscaling, we will  see entire process here and also on the CMD  

• 00:11:58 window. The 1st thing that it is going to do  is it will split the video into chunks. You  

• 00:12:03 see input chunks here. Actually, before this,  it will 1st downscale video. Why? Because our  

• 00:12:09 current configuration requires some downscale.  You see, this is the downscaled version. So it  

• 00:12:14 is going to process this, then split into chunks  like this, then it will start processing each  

• 00:12:20 chunk. As the chunks are processed, they will  start appearing here. This is, by the way,  

• 00:12:25 fully also copying the audios of the videos. So  you won't have any issues with the audio as well,  

• 00:12:32 and the chunks will appear like this. So you will  be able to select each chunk, whichever chunk you  

• 00:12:37 want, and see them. Currently, since we closed  this instance, it is not here. And let's look  

• 00:12:43 at the VRAM usage. The VRAM usage is like 16  gigabytes right now. Lesser than 16 gigabytes,  

• 00:12:48 and we are upscaling into whopping over 4K. This  is like 8 megapixel. So this is a very big upscale  

• 00:12:56 and it is really fast upscale on my local computer  running fully locally. You can see the VRAM usage,  

• 00:13:03 it is fully optimized. With 16 gigabytes VRAM  GPU, you can upscale into this resolution. If  

• 00:13:09 you don't have such powerful GPU, you can  do this 2x or you can reduce this cap.

• 00:13:14 So what other tricks there are for SeedVR2?  There aren't much other tricks. You can reduce  

• 00:13:19 the sparse ratio or you can reduce the frame  chunk size to like 32. So it will use lesser  

• 00:13:25 VRAM. You can also reduce these things, but I  don't recommend them. Just play with the chunk  

• 00:13:30 size and see how it is doing. You can also change  the tile size and tile overlap, I don't recommend  

• 00:13:36 though. Okay, what other features our application  has? Let's also cancel this. You see, I will click  

• 00:13:41 this and cancel, and it will immediately cancel  it, and you will see that now 0 VRAM usage and 0  

• 00:13:47 RAM usage. Okay, we have health check. When you  run health check, it will show you whether you  

• 00:13:53 did set up it accurately or not. We have global  settings where you can set the default outputs  

• 00:13:58 folder and default temporary folder. You can also  set your theme like light or dark. We have fully  

• 00:14:03 working preset system that you can save and load  at every tab, and it is saving and loading every  

• 00:14:10 configuration on every tab that we have. We have  full queue system. You can queue multiple jobs,  

• 00:14:16 just click upscale and it will automatically  queue them. We have face restoration, RIFE  

• 00:14:20 frame interpolation, we have image based upscales,  we have resolution scene split. You can set some  

• 00:14:27 options here about the scene split. And in the  output and comparison, you can set also the output  

• 00:14:34 of the video. Let's refresh this a little bit.  Yes, I have reported some Gradio slowdowns and  

• 00:14:41 the team is working. Hopefully, the application  will get much faster and much smoother. You can  

• 00:14:46 enable global RIFE here, you can set the FPS  increase here. In the output format, you can  

• 00:14:51 change the output format here. In the video  output format, you can set the SeedVR2 backend  

• 00:14:59 like this 10 bit encoding. You can set your video  codec, this is for SeedVR2 H265. You can set some  

• 00:15:07 tuning of H265 like grain. So we have a lot  of options here like the quality, audio codec,  

• 00:15:12 whatever you want. And we have comparison display.  This is automatically generating comparison of the  

• 00:15:18 upscaled videos that you will see in your outputs  folder. Let me show you 1 of them. So for example,  

• 00:15:24 this is a fully completed video. You see that  there is FlashVSR full example comparison. When  

• 00:15:29 you open this, it will show you the original and  the upscaled version right away like this. This  

• 00:15:34 is automatically generated. We are also saving all  the metadata like run summary or the run metadata  

• 00:15:42 or the run context, you will see everything.  We are also saving the processed chunks,  

• 00:15:48 you can see them right away, the input chunks.  If your browser crashes, if the Gradio crashes,  

• 00:15:53 it is not an issue, everything will  be still saved in the outputs folder.

• 00:15:58 We have direct video compare which I used to  generate those videos. With direct video compare,  

• 00:16:04 you can upload the videos and set the custom  titles like the 1st video or 2nd video,  

• 00:16:12 whatever you want, you can set them in here. You  can set the label font size. You can also generate  

• 00:16:17 animated slider, this is taking a lot of time,  and you can have an animated slider comparison  

• 00:16:23 like this. This took huge time to code, believe  me, this was a very hard job, very hard task.  

• 00:16:29 It is fully working. We have videos comparison  slider. So the videos comparison slider is very  

• 00:16:34 cool. You can upload multiple videos like test  1, test 1 SECourses, test 1 Topaz AI, and it  

• 00:16:39 will let you quickly see them. For example, test  1 is the original, this is the SECourses upscaled,  

• 00:16:45 and you can play like this. You can play with this  slider and see the quality difference. You see the  

• 00:16:50 left 1 is original, the right 1 is SECourses  upscaled. You can switch to Topaz AI and click  

• 00:16:57 play, and you see the left 1 is Topaz AI, the  right 1 is the SECourses upscaled, and you can  

• 00:17:03 see the quality difference. We can also sync if  it is not synced. Yes, you see there is a massive  

• 00:17:08 quality difference between 2 upscaled videos.  You can see the details difference. I mean,  

• 00:17:14 the SECourses is adding so much difference.  It is not fully synced at the moment. Yeah,  

• 00:17:19 now better. You see, amazing difference. The  SECourses FlashVSR is adding amazing differences  

• 00:17:26 and it is fully optimized. I mean, let me show  you the default settings VRAM usage. 1st of all,  

• 00:17:33 I recommend you to try default. So  let's refresh and let's upload a video,  

• 00:17:38 test 1. I am not changing anything and let's just  click upscale and let's see how much VRAM it will  

• 00:17:44 use. It is going to upscale it into HD resolution  1920. So it will 1st downscale and then upscale.  

• 00:17:52 So it is using 12 gigabytes of VRAM right now, and  I can reduce this even further. How can I do that?  

• 00:18:00 I can reduce... Oh, by the way, this is SeedVR2,  not the FlashVSR. The SeedVR2 is more demanding.  

• 00:18:08 So 1st I recommend you to use FlashVSR, then  the SeedVR2. So I will just cancel this. Okay,  

• 00:18:15 it should be cancelled. Yes. Let's go to FlashVSR.  Let's upload our video, it is all default, and  

• 00:18:22 start upscale. Let's see the difference and let's  see the VRAM usage. It is going to upscale the  

• 00:18:28 video into HD resolution. Since my original input  resolution was not very low, I would upscale this  

• 00:18:35 into 2x with max resolution normally, but this  is a demonstration to show you the VRAM usage. So  

• 00:18:42 currently it is using like 16 gigabytes of VRAM.  Actually, it didn't make much difference. Oh wait,  

• 00:18:49 maybe it is not using. So let's cancel this.  Let's just refresh to get the default. Okay,  

• 00:18:58 test 1. Let's make this 2x upscale. Okay, let's  verify. Yeah, looking good. Let's see the VRAM  

• 00:19:07 usage. Okay, let's upscale. So play with the  parameters, see which 1 is working best for your  

• 00:19:13 case. Try to not use shared VRAM because if you  start using shared VRAM, it will become too slow.  

• 00:19:19 Sometimes it doesn't give out of VRAM memory, it  is just using the shared VRAM of the Nvidia. Okay,  

• 00:19:25 now we are going to see the VRAM usage.  Yes, it is using under 8 gigabytes of VRAM  

• 00:19:30 GPU and it is upscaling amazingly fast into HD  resolution. So this is really pretty amazing.

• 00:19:38 And now I will show you some things about  the upcoming Trellis. I am working hugely on  

• 00:19:44 Trellis application. I have added UniRig  rigging to our Trellis application. So  

• 00:19:51 that now you will be able to generate  skeletons to your characters like this,  

• 00:19:56 you will be able to use them in any 3D. It is  not perfected yet, but I am working on improving.  

• 00:20:04 These errors you see will be all gone. I am  adding massive improvements to the application,  

• 00:20:09 I am working on it. Adding this 3D rigging, model  rigging took my huge time. I am using UniRig and  

• 00:20:16 I had to compile so many different libraries.  This is working on CUDA 13 and Torch 2.9.1,  

• 00:20:23 supporting all the GPUs out there.  Currently, I am using RTX 5090,  

• 00:20:27 but working with every GPU and working really fast  and optimized. These took my huge time. I mean,  

• 00:20:33 3D world requires so many custom libraries,  custom open source projects, GitHub repositories,  

• 00:20:40 and none of them is ready to be used by the newer  GPUs, they are all outdated. So I had to compile  

• 00:20:46 every 1 of them and it is working amazing right  now. We are saving all the formats directly. So  

• 00:20:53 let's open outputs folder to show you. You see,  we are saving skeleton FBX, preview GLB, skinned  

• 00:21:01 FBX. It is doing everything with the texture as  well. And as I said, it will get much better. I am  

• 00:21:08 also going to add multiple angle input hopefully,  like front, right, left, and backwards. So let's  

• 00:21:15 import for example, let's import this 1. I am not  a 3D expert, so I can't tell you much, but okay,  

• 00:21:25 this is the FBX. This has the pose armature  bone. Okay, let's import the actual skin 1. When  

• 00:21:30 I delete this, you can see it like this. So let's  add another input like this 1. Okay, import. Okay,  

• 00:21:39 this is like this 1. I don't know what exactly  it is, but it is like this as you can see. And  

• 00:21:45 we have textured preview, bones and skins. You  see, it is showing everything. Okay, let's see  

• 00:21:51 the rendered video. Yes, like this, you see. And  it will get much better hopefully, I am working  

• 00:21:55 on it. And let's see there is skeleton preview.  Let's see this 1. Okay, you see? It's amazing  

• 00:22:03 in my opinion. And you can see the generated  skeleton, the rig, the body. Yeah, pretty amazing.

• 00:22:11 So I am working on it. Hopefully, it will  get much better. It is not updated yet,  

• 00:22:15 but soon. The link of the Trellis 2 and the  upscaler will be in the description of the  

• 00:22:20 video. If you have any questions, just message me.  I am adding new features, new request features,  

• 00:22:25 like someone requested me to 1st upscale 2x with  SeedVR2, then upscale 2x with FlashVSR. So I am  

• 00:22:32 going to hopefully add that too. You will be  able to multiple stage upscale. It will be  

• 00:22:38 based on the queue system, it will be added here.  This page will get much more improved. So use it,  

• 00:22:44 test it, report me any errors, report me any  request that you have. Hopefully, I will make  

• 00:22:48 it much better. This will be the ultimate video  and image upscaler. Hopefully, I will add newer  

• 00:22:53 models to here as well. I may even add the SUPIR  into this application as well, if it be easy and  

• 00:23:01 possible. So I am working on this application.  This will be the feature 1 for all application,  

• 00:23:07 much better than Topaz AI or any alternatives. Any  newer model that comes will be also implemented  

• 00:23:14 into this application. I am investing huge time  into this application. And the Trellis 2 will be  

• 00:23:19 the ultimate application for image to 3D with lots  of features like rigging or like multiple angles  

• 00:23:26 input or whatever there will be. Okay, hopefully  see you later in another amazing tutorial video.