Search options (Page 163 of 236)

lkk_12342001
Thank you for your help to make SVP more international.

I see both logs are identical from view to settings folder and smooth properties for the file.
I see one little mistake in english log:

SVP-manager must be close but it continue its work. It is strange. But it is not cause of performance difference between users.

Check GPU-acceleration (OpenCL): true in both cases.
And look at filters list in both cases. You must find differences there. I think it is decoder and renderer.

We have some international improvements in SVP 3.1.3 beta and we need to test it.
Would you like to become international SVP beta tester?

PsykoTenshi
Thank you. I saw the same picture with "hill" at memory use graphic. It is dynamic subtitles. It was not enough time to make some tests with it.
I will compare different subtitle filters and renderers to find better.

TCmullet
I think that you have very long chain of video recording from net:

And I think you need more short chain like this:

No decode. No encode. No Losses and no CPU load. Save stream "As Is".
What will you say?

you saying that internally you change it to 120fps and (also internally) decimate by 2?
Yes. It processed internally. And it is optimized. It works without rendering decimated frames.

Does InterFrame come up with whatever is needed to allow any combination of input and output FPS?
Yes. You can use any combinations of these FPS. But you can get maximum quality and minimum CPU load if you using multiply FPSs: from 25 to 50 fps, from 30 to 60 fps, from 23.976 to 47.952 fps.
Or fraction multiply with low devider: 24 * (5:2) = 60, 23.976 * (5:2) = 59.94.

In first case all source frames are saved in the output video, in second case only - only part of source frames will be saved.

PsykoTenshi
I have never seen such strange memory usage line.
Can you share this file to me or part of it what shows such behavior?
I will take a look

I looked at one of hard cases: AVC Hi10p 4:4:4 (50fps but only 720p)
H.264-444-10bit-sample (1,4 MB)
It showed 30% of CPU usage with ffdShow decoder and 25% CPU usage with LAV decoder. Hi10p decode now is software only.
CPU: Core i5-2300 2,8 GHz

theosoph
Сравнить оба уплавнителя (тот что в мониторе и SVP) и решить для себя, имеет ли смысл.
Что за монитор?

%username%
Да, ситуация получила развитие как раз по твоему сценарию. Переставлять винду sturm уже не будет, т.к. никакой нормальный пользователь не будет копаться со старым железом, если приехало новое и работает с нуля лучше старого. Поэтому мы никогда не узнаем реальную причину потрескиваний X-Fi звука на AMD

---
Раз так волнует судьба SVP и есть интерес к развитию пакета, то может того... К нам в бета-тестеры?
Отправил подробности в личном сообщении.

In usual case 24 fps is better than 25 fps to be interpolated to 60.
You are almost right. 24 fps encreased to 60 fps by added several interframes. Formula is 5:2. That mean to add 4 new frames and decimate 1 from each 2 sorce frames.

But I tell about real 24 fps.

Your fps after capturing is not real. It is fake fps. No guarantee that all frames are at it's own positions. You need not to capture. You need to save video stream with no changes or with minimal changes.

gaunt
плавность при перемотке ...меня , например, раздражет отключение плавности при *2 режиме : доступном в хомесинема . Ибо часто пользуюсь для быстрого просмотра всяких рипов дискавери и т.п. Но , блин , чтобы раздражало других - нужно прикупить этим другим камень ...баксов за 200-300
Надо всего лишь рассказать об этом разработчикам.
Почему я читаю это впервые и здесь, а не в закрытом бета-форуме?

TCmullet
In most cases, each frame (in my source) is different than the prior one
That's not enough. You need equidistant frames.

Good video is constant frame rate video with one time-delta between frames. Your video has not one time-delta. They are various at each frame.
If source have 25 fps then no way to get 15 fps without harm to time-delta. Such irregularity are present in your video it not at one place, it is everywhere.

no matter how far apart they are.
You are right. In your case you can get some result. But to get good result you need good source. Your source is bad.

I realize that if a framerate was really low, so low that any wild actions could occur between 2 snapped frames, then of course no lost action can be truly created.
Yes. You are right.

But if an arm moves from, say, pointing down 45 degrees from horizontal, then the next frame shows it pointing upward 45 degrees, I'd think InterFrame would be smart enough to generate an inter-frame with the arm pointing 0 degrees
No. Interframe can get 0 degrees arm at out only if sources has -5 and 5 degrees. 45 degrees are too much for this task. No rotation predicted. Only global shift and local shifts.

why does the ball, when appearing very discretely (round and clearly formed) sometimes get interpolated as 2 balls in one interpolated frame?
Because ball is small and his moves are large.

I have no idea how to help you. You need a lot of hard hand work frame-by-frame to reconstruct dropped original frames.
Such big intervals between frames for sports is very hard to any interpolator. Too small details, too large moves and differencies between frames.

Sorry. I have no time to do your job with your video. I must to work at new SVP version. Maybe after SVP release I will find a time to help you more.

TCmullet
I had a look at your video. It is bad. It has framedrops. Motions are not "uniform".

How did you get this video?
15 fps is not standard.