Video Enhance AI v2.1.1

It’s not even a beta yet.

Ok! Please add another display of the time elapsed from the start of encoding, next to the time remaining until the end of encoding. And show the time after the end of encoding how long it took.

1 Like

It seems to me that AI should think like a human. If a person sees an object at a distant distance, then he must also recognize it or guess. Next comes the association, what it might look like. The advantage of AI over humans is that it can quickly operate with large amounts of information and look into the future plots of the entire film, return and create a model in space, remember it in place of space and scale it depending on the scene and angle. Modern games are designed this way when there is a map of all buildings, objects and trees, and based on the position in space, everything is modeled in the game very quickly. On the video, you can do the same, the main thing is to recognize the object, not from the first frame, but by viewing the entire scene until the end of the plot. Or maybe in the next plot they will show a close plan and there will be all the details for building a 3D model. You can also recognize the 3D perspective of the video and apply different filters for near and far objects in relation to this. And having the previous and subsequent frames of the video, you can restore the damaged background around the moving objects, which, like a logo, could be removed. Yes, I have already seen how this is accidentally modeled in topaz, and basically, now an unrecognized object can turn out to be anything from the first frame and migrate to another object as you approach it or change the camera position. Therefore, I suggested, when changing the plot, to analyze and process the video not from the first frame, but at least from 10 to the first and then from the restored first to the tenth and further until the end of the plot. And by the way, black frames could also be recognized and apply a quick filter to them for fast encoding, but now it takes the same time for all the complexities of the image.

Accidentally changing the grain values while scrolling down to the video format area has been the bane of my existence lately…

I checked on that. They release at the exact same time. 20H2 has been available since October 2020 for Home, Pro, and Enterprise. I suspect you have a “deferral tool” of some sort on your system, or a registry hack that pushes updates out to their limit.

Keep in mind, even if it was possible, your method still wouldn’t work for people that only appear in the background of a video (never in the foreground). So it would look more than weird if a person was in a background crowd and then walked forward towards the camera. Because he would have a nice clear face while everybody around him would look like zombies. :laughing:

I am pretty sure that version 1.5.3 to 1.7.1 quite good and was working on my PC (CPU: I7-8700, GPU: MSI RX 5700 XT Gaming, RAM: 32GB-4000mhz, Samsung 980 Pro nvme). Even increase the GPU slider up to 80 was fine for the software.

When I want to try 2.1.1 it starts crashing my PC in 3-5mins no matter what settings/video file type I use. After my 2 day marathon to figured it out what is wrong, I reinstall fresh Windows to both of my machines.Then tried all af the previous versions (v1.5.3 / v1.6.0 / v1.6.1 / v1.8.1 / v1.8.2 / v1.9.0 / v2.0.0 / v2.1.1). All of them crahsing.

I changed GPU to another slot and I tried another video file. I cleaned the video file with Premiere Pro and export as xmf. As I wanted the color correction after VEAI enhance.
The xmf file is 64GB and 1280x720. I loaded to VEAI and choose to HD 13920x1080 mov file. My PC crashes after 1/4 of the file. BTW: When I was looking through web found so many users have problem with crashes. I came accross an review about VEAI which saying is the software file size limitation is 10GB. But review was quite some time ago. I don’t know it still has that limitation or not.

I watched of all process, tried with smaller file size and figured out VEAI not write data to disk during process. I think All data kept in either memory or temp file. Don’t know it is related to it.
I am an an IT expert but I am out of options. Same PC, same specs.GPU checked, PSU checked (voltages seems ok), CPU checked, slots ok. Other hungry resource apps (4 games working well with 1920x1080 resolution). If the software was working with the version 1.7.1 prevously why doesn’t work now. Only two things changed/added to PC: Windows 10 Updates on March and April. Seems that the software not compatible for the latest updates of Windows 10 (I don’t know). Seems like VEAI cannot handle for using the system resources well. Because all of my Adobe programs, my resource hungry games working like a charm.

2 Likes

Some of your suggestions here for Gaia we already asked Topaz months ago. In response, they said it was not possible to adapt Gaia in such a way, however what they came up with as a ‘solution’ is the Theia model. No, it did not come even close, but that’s how Theia came to be. :blush:

In the months that followed, Topaz tried (and failed) to improve Gaia by any significant amount. They then focused most of their efforts into developing Artemis, and later Dione, into what we have today.

1 Like

I agree! This requires a well-trained AI. This requires the right algorithms. AI works precisely according to algorithms that were written by a person and not otherwise. To do this, if the background people no longer appear in the frame, and they are no longer in close-up, then they are simply assigned a face similar to the silhouette of a face from the template database, even if it does not look like anyone else, but it is better than a zombie with horns and three noses and eyes on the cheeks. Or at least to make it clear that this is a human head, and nothing else and ugly. It seems to me that this is possible, but only super computers can do it, because a personal computer will take too long to render such videos, it’s like creating realistic plots for the Avatar movie.

I have the same story, but it only happens with one video card AMD PULSE RX 560 2Gb and on different computers with different hardware. I can’t understand what is happening, but there was no such thing before, but after some updates and after 2.1.1 it became like a program crash with an error and file corruption. Also, reinstalling old drivers and programs does not help. The video card memory and graphics test pass many tests without problems. Also, encoding and filters in other graphics programs with a heavy load on the video card - everything is fine. If insert an older video card R7 260X 1Gb, then everything is fine and without problems. Also, it happened after the next Windows update, and it looks like something got stuck somewhere in the registries and spoils everything. I also noticed that with 1GB of memory you can encode only if you make the settings in the settings: Max VRAM Usage - Low; and turn on Reduce Machine Load and Disable Model Downloading.

Yeah, if the AI could recognize that the ‘blob’ in the background was a person, male or female, short hair or long hair, skinny or fat (etc), then it could pull the appropriate digital ‘doppelgänger’ from a database. In theory, it could work.

1 Like

Please add more selectable modes of automatic computer actions at the end of encoding: Sleep mode; Hibernation mode; Complete shutdown mode with the closure of all programs. This is very necessary, because the process of work is long, if the process ends at night, then the computer itself turns off according to the selected mode, if this is chosen. Thank you!

I have invented and tested a simple feature that improves Topaz Video Enhance.
If you use it, I would like to take credit under my real name. Let me know.

I reverse the order of frames to effectively de-noise the first frame through look ahead/back:
Frame No. 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 2, 3, 4, 5, 6, 7, ETC.
The reversed frames are removed after processing.

If I understand correctly, Topaz Video Enhance processes pixel aspect before upscale, which would beg the question why you would not include pixel aspect in upscale, processing square pixels.

Before Lanczos reduction of the other dimension to match correct pixel aspect.

This could be a versatile feature, that might also include specifying pixel aspect to work with AVI, or override file metadata. In its simplest form, reduce chosen dimension.

I de-noise in SAR because we only reduce quality and add more noise by changing this to DAR first.
I would like the ability to disable any facial recognition features in CG-HQ.

Obviously, the use of intermediate codec like Lagarith 32 bit would be ideal for workflow.

Any news on support for newer Nvidia drivers? There’s already a new driver released, so we’re 2 driver versions behind on support now ):

1 Like

Hello, good news
NVIDIA Studio Driver

Version : 462.31
Date : 2021.4.14
OS : Windows 10 64-bit

NVIDIA Studio Drivers provide artists, creators, and 3D developers the best performance and reliability when working with creative applications. To achieve the highest level of reliability, Studio Drivers undergo extensive testing against multi-app creator workflows and multiple revisions of the top creative applications from Adobe to Autodesk and beyond.

Applications

New AI features in top creative apps, running on NVIDIA RTX GPUs, are changing - and accelerating - the way we create. The April NVIDIA Studio Driver provides optimal support for the latest AI-powered features in creative applications including NVIDIA Omniverse, Topaz Denoise AI, Topaz Sharpen AI, Notch, and OBS Studio.

In addition, GeForce Experience now optimizes creative applications like Blackmagic Design’s DaVinci Resolve and Adobe Lightroom by automatically enabling GPU acceleration in these apps, ensuring artists are using the maximum GPU acceleration available, just by clicking a single button. Initially starting with 34 supported apps, this list will continue to grow over time as more creative applications are added.

Hope this help

Try deleting or moving temporarily to a different folder all model files from the system folder (C: \ Users \ **** \ AppData \ Roaming \ Topaz Labs LLC \ Video Enhance AI \ models). For the selected model: amq-v12-gnet-fp32-288x288-2x-ox (with the value -fp32-) and leave all files with (-fp16-) and deny access to download them in the program settings. I stopped crashing, but it is encoded twice as slow. Check how it is with you and write.

Hello,

Thank you for your respond.
I checked the folder you referred to. But the files under “\ AppData \ Roaming \ Topaz Labs LLC \ Video Enhance AI \ models” all are has -fp16- value.
None of the has -fp32- value.

Please see the attached picture
Imgur

And today I tried something new;
I changed Windows Power options to “High Performance”. (was Balanced)

Then unhide the Processor Power Management > “Processor Performance Boost Mode” option with the registry hack under Power Options > Advanced Settings.

Disabled “Processor Performance Boost Mode” (was aggressive on default)

Run the software. PC now crashed after 20-30 mins instead of 3mins.

I returned BIOS to default settings = crashed again in 30 mins.
I wrote to support service. They asked me to send the logs for investigating.

Are there any new model that comes close to CG HQ for low res/quality clips pre-cleaned with Neat Video? With older versions, CG HQ was miles ahead.

I find that blocking needs to be broken up by neat video, in order to be recongnised and removed by CG HQ. The program cannot deal with full blocking squares!

Ok! (-fp32-) is downloaded from the Internet on its own, if not prohibited in the settings. At the same time, the more the memory level regulator is set in the program settings, the higher the block values ​​are for the modules for the used model. This can be seen in real time when coding, how files are uploaded to this folder. Files of models with blocks (480x384 … 576х448 …) (- fp32-) can be loaded.
Also, I installed the card in the motherboard with AMD 20.12.1 drivers and Windows 20H2 19042.508 with updates to 2020.12. program 2.1.1 kept crash. Then I removed them (- fp32-) and it works now. But program 2.1.1 also didn’t work with files (-fp16-) on a system with the latest updates and the latest driver. Today I will check it sequentially, after which it will crash again if I install the latest drivers and updates. Or maybe the direct x (d3dx9_43.dll) files that I additionally installed for some programs. Also, the Reduse Machine Load setting strongly influenced the operation of the system itself, and sometimes if you turn it off, the program immediately reported a lack of memory with a 1 GB AMD card, and if you turn on this function, the program worked. And also, with AMD R7 260X 1GB cards, the 2.1.1 program works with any files and updates.

Will we ever see Gaia CG restored to the quality found in 1.6.1? I assumed that since this is no longer listed in the issues that it was fixed in 1.9.0, but after switching back to 1.6.1, I can see that it is still the superior Gaia GC model.