Topaz Video AI v3.0.4

Topaz Video AI Releases
196

This is what I’ve been doing on determining the settings to use for parameters.

I created a spreadsheet that has 26 Columns and and 6 rows.

The Rows are for 6 settings in Parameters.

The Columns are for 25 testings at different choices in the movie that I choose at random and do an analyze for settings.

I put the 25 random samplings for testing what the parameters should be and the 26th line averages them.

I use the average of the 25 settings to process the enhancements and this has worked out really nice for the video enhancements and output qualities.

This should be a simple thing to do to add to this program and depending on what the developers feel is a good sampling (25? 50?) should result in a good tool for people to use and should be a fairly quick process of a few minutes at the most.

Is this a good idea and do you think it should be added into the parameters options as another choice (currently Auto, Manual, etc…) maybe labeled Samplings 50 that when you run it will do what I do manually.

Do 50 random samplings at various points and use the average as the value to process enhancements for the video.

Thanks!

Mike

4 Likes
197

Not a bad idea at all.
I have also noticed that different settings can be necessary at different points.

2 Likes
198

This is one of the first things I noticed as a newbie to this product (purchased 10/30/22).

The sampling is very limited right now for estimating parameters and if you happen to hit upon an odd segment of the movie (unusually dark, indoors vs outdoors, action scene vs. “normal”, etc…) the settings it uses for parameters is not representative of the movie as a whole.

Mike

200

In Topaz You don’t have any control on coding process.
I always set Topaz for max HQ then coding in VirtualDud2 as I want coding (CRF, b-frames, Film preset, etc.).

1 Like
201

The performance in 3 vs 2.6 is really really bad. The estimated time to process a short clip has gone for 8-10mins to hours for me, using an AMD RX570 8GB video card.
Preview is not working properly. When the preview finally appears and you try to A/B it, the frames are rarely the same, which makes accurate comparison impossible.

The trim feature is a big step backwards from the one in 2.6. It is much less practical to use. It was much better when you could punch in, and punch out the beginning and the end of the trim. The ability to ‘apply’ the trim is useful.
I think that a better implementation of trim would be the ability to create multiple regions by punching in, and punching out, then applying, with the ability to process and export multiple regions as a single video file.

I think that the crop functionality is really good and appreciate that feature a lot.

4 Likes
202

Видео продолжительностью в 17 секунд обрабатывается 22 минуты… Это по вашему нормально? При этом каждый раз на половине обработки на N% сбрасывается и обработка начинается с самого начала. Я чувствую себя бета-тестером, которого вы наняли, чтобы я искал баги… Только вот чтобы найти баги тут не нужно далеко ходить, достаточно добавить в обработку более 20 роликов разного разрешения. Баги которые я нашел:

  1. При экспорте визуально ничего не происходит, даже нет ощущения что кнопка нажалась.
  2. Чтобы увидеть превью или экспорт в столбце состояния в центре нужно переключить вкладки, новые строки появляются только после переключения вкладок.
  3. Не обрабатывается много видео. Когда выделяю много видео и нажимаю экспорт (визуально ничего не происходит) обрабатывается только 1 видео. Находимо ждать пока закончится обработка одного ролика и нажать на экспорт еще раз, если нажать экспорт сразу несколько раз программа закроется.
  4. 17 секундное видео обрабатывается 22 минуты на RTX 3090. Таких проблем не было в версии 2.6.4.
    Я пользуюсь программой 1 час, 1 час назад я её купил… Вопрос к вам, почему вы продаете сырой продукт? Тут даже стандартные функции, которые вы сами заложили не работают в полной мере. Если вы мне дали версию 2.6.4 с её старым функционалом я был был больше рад, чем от того что вы даёте мне сейчас. Я чувствую себя обманутым. Программа стала хуже чем была раньше про производительности, по качеству обработки, по рабочим функциям, разве что стилистика стала лучше и на этом всё. Я купил вашу программу чтобы обрабатывать много видео одним нажатием кнопки. Сейчас я сижу и жду обработки каждого ролика чтобы нажать кнопку экспорта для обработки следующего…
2 Likes
203

On 100% is bad too. I zoomed to 200% to better see error.

204

y7moonT0rnad0 You make several very good points, and I agree with them all.

Most have been made by other people previously but the developers should translate your post - and fix all these things. If this will save anyone a little time, here is the google translation:

y7moonT0rnad0

A video of 17 seconds is processed for 22 minutes … Is this normal for you? In this case, each time at half processing, N% is reset and processing starts from the very beginning. I feel like a beta tester that you hired to look for bugs… But you don’t have to go far to find bugs, just add more than 20 videos of different resolutions to the processing. Bugs I found:

When exporting, visually nothing happens, there is not even a feeling that the button was pressed.

To see the preview or export in the status column in the center, you need to switch tabs, new lines appear only after switching tabs.

Not many videos are being processed. When I select a lot of videos and click export (visually nothing happens), only 1 video is processed. We find to wait until the processing of one video is completed and click on the export again, if you click the export several times at once, the program will close.

A 17 second video takes 22 minutes to process on the RTX 3090. There were no such issues in version 2.6.4.

I have been using the program for 1 hour, I bought it 1 hour ago… Question for you, why are you selling a raw product? Here, even the standard functions that you yourself have laid down do not work fully. If you gave me version 2.6.4 with its old functionality, I was more pleased than what you give me now. I feel cheated.

The program has become worse than it was before in terms of performance, in terms of processing quality, in terms of working functions, except that the style has become better and that’s it. I bought your program to process many videos at the click of a button. Now I sit and wait for each video to be processed so that I can hit the export button to process the next one …

3 Likes
206

Y7moon,

I agree with you on a few of these points, but I also disagree in the main, even though you are echoing many of the complaints being given by others. Yes, the new version is slower in many ways that v2.x was. I believe that they will rectify the speed problems before too long. There are a lot of other small glitches that need to be ironed out first.

I have also (slowly) come to the conclusion that TVAI 3.x is a much different tool than VEAI version 2.x was. - That means the techniques we became accustomed to when using v2.x will not work as well if we try to apply exactly the same methodology to version TVAI v3. .

After coming to that realization, I started approaching my use of TVAI 3 as if it is a new tool and as a craftsman, I must learn the best way to use its potentials to do even finer work. - After coming to that conclusion, and devising alternate methods, I started getting superior results from TVAI 3. (Yes, despite the defects in the new release!)

I will continue to keep the forum and developers advised of the numerous bugs I am encountering, but I’m also benefitting from far better output; it’s mainly a matter of adopting new methodologies.

:nerd_face: (Also, I’m very certain there are many participants in this forum who will disagree…)

207

Was TVAI or TVEA designed to down-convert video? It seems that you are using the wrong tool for your project (or at least 1/2 of your project).

I am not aware of using AI technology to down-convert video. I am not even sure that going from 30p to 60p uses AI technology as well. I get incredible 30p to 60p conversions using Teranex hardware that does not use AI.

208

No more export to JPEG? It was the only suitable solution to pause/restore processing.
Other image formats are not compressed. To much space to use…

1 Like
209

Дело ведь не в том что вам нужно в первую очередь, а в том что мне нужно в первую очередь. По этому чьё-то согласие или несогласие тут не требуется. Я купил продукт, чтобы выполнять какие-то задачи, не тратя на это слишком много времени. На текущий момент мне будет достаточно исправление многопоточной работы без вылетов и ошибок. Всё что мне требуется, работало ранее. Я не прошу придумать что-то новое. Даже если об этих проблемах неоднократно говорили… Я человек который купил продукт в надежде что я сразу смогу в нем работать без проблем, а тут выясняется что это сырой продукт, который имеет версию 3.0.X… Нужно было начать с 0.1.X, чтобы люди понимали что они покупают. Видимо это для всех считается нормальным выпустить версию 1.0+, собирать деньги и кормить людей обещаниями. Простите что я этого понять никак не могу. Почему разработчики софта начали вести себя как разработчики игр? На примере релиза Cyberpunk 2077 нам тоже нужно будет год ждать чтобы то что должно было работать с релиза заработало? Извините меня за то что я не доволен чем-то.

210

Hi
Does anyony know when previews will be fixed?

Thanks

211

The jump from VEAI 2.x to TVAI 3.x was major. Everything was rewritten. And, not only was the older functionality replaced with newer versions, they were chained together differently. This makes possible the use of multiple filters in a single pass. This required rethinking the roles of each filter and how they relate to each other.

It takes a bit of floundering around to become accustomed to the way these capabilities can be teamed up to give maximum results with the least amount of effort. Of course the processes may be a little slower, for the present, but when properly configured, the results are well worth the wait.

As far as the “raw” condition of this “released” product. A lot of the beta-testers were there to witness the evolution of v3.x beta from a bunch of ‘ideas’ and into a fairly well integrated system. But most of us thought that the production release was done a bit too early.

As I said in my previous reply, there are still a lot of glitches, (or bugs,) that need smoothing out. TVAI 3 has been a very complex developmental undertaking and the amount of software, and layers of software involved is considerable. There are bound to be adjustments and fixes for a while to come. However, despite this a lot of us are learning how to work around some of the problems for now, so we can take advantage of the advanced capabilities, and (of course) report any defects or make suggestions that may make the software easier to use and more reliable. - That is a big part of why this forum exists.

One thing I’ve discovered. is that there are some lower-level issues that can crop up during previews, which may cause the final export job to fail. My workaround for that is to save my settings in a Preset. Close TVAI and do a full reboot. Following which, I load my source, recall the saved preset, and export immediately. - I am certain that in the future, the problem(s) causing this will go away.

By the way, it is not uncommon for the primary release of any new and complex body of software to have some problems. Many of these are not detected until the software has been widely distributed. - It is impossible to anticipate all possible problems in a testbed environment, especially when the users/tester/programmers are already familiar with using it. And as I said before, TVAI 3.x is not just an upgraded version of VEAI 2.x, but completely rewritten.

I have a very fast system with an overclocked RXT 3090 in it. And I admit that I would like TVAI to run enhancements faster. I’m moving to an RTX 4090 soon, but from what I have heard even the more powerful GPU will not perform miracles. Even so, I am confident that the problems will go away and the processing will get faster. Until then I am using it and being patient until things improve.

As a woman from Montral, Canada once said to me “Paris was not built in a day.”

As my first team manager in a software development project told me "It’s the first hundred years that’s the hardest.

:thinking: :nerd_face:

3 Likes
212

Just to let you know that if you’re unhappy with the upgrade or the software you bought, you can write to the support and ask for a refund. there is a 30 days Policy refund.
if you still want to keep it, you can then join us, and wait that the issues will be or maybe fixed (or not ! lol).

213

I was mainly using video enhance to improve the quality of my full 3d renders. It was a real + since we could cut the costs of the renderfarm calculations by half and still have a nice product in the end.
The last version is simply unusable for me. I tried dropping the usual mp4, can’t anymore. Image sequences make the soft crash. Avi files can’t be imported either. 2.6.4 was accepting all formats without any hassle.
Performace is half 2.6.4 when i manage to find a proper format in which to import my video.
Im reverting to 2.6.4

2 Likes
214

Installed 3.0.4 today. First attempt at rendering and the whole app is unresponsive…Not a great start.

2 Likes
215

PD

Have you done any scene splitting with the Enhancement sections? I started trying this with videos that have varied dynamic change within the video. I’m still doing my lossless preprocessing first but I’m finding that I’m getting some good results. It’s so easy to split a video during pre and than put it back together in post. This gives me the ability to use which ever model gives me the best results for a scene rather than hope things like faces don’t get destroyed in some scenes and not others or artifacts are generated in some and not others.

216

Yes, that is the good old-fashioned way to do video editing and enhancement. Proteus, and to a lesser extent the other VAI enhancers are an effort to reduce the need to cut everything up into clips in order to do enhancement.

However, in my opinion, cutting a video into separate clips inside of TVAI gets kind of tedious; the clip selection interface is kind of awkward and there is no methodology in place for saving them in an orderly way. (TVAI is essentially an enhancer, not an editor.) So, when I need to do that, I simply go to a ‘regular’ video editor, make the cuts, and save them in their original format, - That is assuming you have the video already deinterlaced and in a good non-lossy format that both your editor and TVAI can use.

After that, you can enhance the separate parts and reassemble them far more easily.

There is also an ‘easy-peasy’ method. It is time-consuming, (it is more work for your computer and less for you.)

But, I’m not including it in this post. :roll_eyes: :thinking:

217

Oh, forgot to mention. Lately I haven’t had too many problems with faces. I believe that TVAI has improved somewhat, but also my familiarity using TVAI has improved. - Both of these things have made that problem fade away.