Ah, OK.  So I guess I'll have to downscale with a different program unless I directly encode to a format that's non-demanding enough that even at 4k 60fps can be decoded in software via an i5-520M (because the discrete NNvidia NVS 3100M can't do 4k hardware decoding).

EDIt:

Chainik wrote:

nothing can stop you from modifying scripts to add custom Vapoursynth code after the conversion

Except my lack of any sort of coding knowledge! tongue

Well I was told that SVPcode can already downsize a video: http://www.svp-team.com/forum/viewtopic.php?pid=66768

So perhaps the question should be - can that resize function work without downsizing the chroma resolution as well?


EDIT:
Perhaps I should mention that I know that I could easily use SVPcode to only encode the 4k 30fps video into 4k 60fps and then use a different program to then downsize to 1080p with 4:4:4 chroma, but I figured I might as well ask and see if it's possible to do it all in one step and save at least a bit of time (I'm only running a Pentium G3258, not a Threadripper 1950X)

As you may know, a 3840x2160 video with 4:2:0 chroma sub-sampling will have a chroma resolution of 1920x1080.  Since I was going to resize a 4k video down to 1080p anyway, I was wondering if SVPcode could manage to do that without also reducing the chroma resolution?

Now I realize that SVP's interpolation engine itself works at 4:2:0, so the interpolation would have to be processed and applied at the original 4k resolution - that's fine.  I'd just like to know if, after that interpolation process, SVPcode can then downscale the video without reducing the chroma resolution.


I imagine that the processing chain would look something like this:

4k 4:2:0 video -> interpolation -> downscale -> 1080p 4:4:4 video

Chainik wrote:

if it's neither 4 nor 8 cores then I prefer to think it's 4 cores/8 threads big_smile

That's why some people just say "4modules/8threads" since it's actually 100% accurate while avoiding the entire "is it truly 8 cores?" debate

Chainik wrote:

FX is "4 cores/8 threads CPU" too

But it's not...?  It's 4 modules, and each module has 2 integer decoders and a single shared floating point decoder.

Windows identifies FX CPUs as 4cores/8threads just so that the scheduler loads each module with an individual single process before it starts putting two processes on a single module.

81

(4 replies, posted in Using SVP)

Chainik wrote:

just throw it away already

Hey, the netbook was free - I just had to replace the hard drive (stuck a cheap 32GB SSD in it).  Money was a problem until a year ago, at which point Ryzen was too close for me to consider upgrading.  Currently Ryzen mobile is way too close, so upgrading now would be incredibly silly.

Also I do actually have faster laptops including ones with a Core 2 Duo and an AMD E-350 (again, both laptops were free save for needing to replace their hard drives), but the key thing is that netbook for some reason has the best quality screen to a noticeable degree (and I don't mean resolution, but actual colors, viewing angles, and black levels).

Lastly the screen resolution of 600px high is perfect for visual novels (which commonly use 800x600) since running 800x600 on a 1280x800 or 1366x768 laptop results in the text looking like a blurry mess, and using windowed gives me a screen size that's not really any larger than the netbook's 10" screen anyway.  Besides, many such 800x600 visual novels run perfectly fine even with the Atom underclocked to 990MHz let alone at its stock 1.6GHz, and obviously the portability, passive cooling, and flicker-free screen (you can test via a fidget spinner) are quite nice when it comes to any sort of lengthy reading.

Chainik wrote:

FX-8350 is equal to or even better than i7-2600k wink

Which is still pretty sad when you consider that a 2600k is only a 4core/8thread CPU.

At least it makes Ryzen a without-a-doubt upgrade over the older FX series - even a lowly Ryzen 3 is competitive. tongue

83

(4 replies, posted in Using SVP)

Simply put, is there anyway for SVPcode to resize the video as well as interpolate it?

I have a weak netbook with a 1024x600 screen powered by a 1st gen Atom and Intel 945GMA integrated graphics, so it has no h.264 hardware decoding.  This is key as 1280x720 @ 60fps video is simply too much for the little CPU to handle, but 1024x576 @ 60fps in a less-demanding format (like Xvid) is lightweight enough to run just fine.


For reference, this aforementioned netbook is by far not my primary PC (that'd be my overclocked G3258 desktop), but I'm at least holding out for Ryzen mobile before I decide to buy any new ultraportable laptop or similar

84

(8 replies, posted in Using SVP)

madvr has a setting to specify resolutions and refresh rates.

Make sure 1080p24 or similar is not listed in that setting, and remove it if it is.  You may also need to change the setting to say something like 1080p60 instead.

Try with a fresh copy of MPC-HC Portable: https://portableapps.com/apps/music_vid … c-portable

Follow the instructions here to configure it for SVP: https://www.svp-team.com/wiki/SVP:MPC-HC



kadigan.ksb wrote:

Mac Pro from 2010 -- it has dual Xeon quads, one of which should have slightly worse performance than my FX-8350

You sure about that?  An FX-8350 has worse performance per-GHz than a Core 2 Quad from 2007.

That's why Ryzen's 50% per-GHz performance increase over FX was such a big deal.



kadigan.ksb wrote:

when I run SVP on that machine, it barely even registers at maximum settings (<2% CPU use

...that to me really sounds like SVP isn't actually doing anything even though the program is open.  Are you able to see an actual visible difference?  Because with "maximum settings" the difference should be really obvious between SVP disabled and SVP enabled.

86

(423 replies, posted in Using SVP)

Isn't Optimus support a total PITA on Linux?  Can't help but think that might be related...

87

(423 replies, posted in Using SVP)

Chainik wrote:

because it doesn't work with a half of display managers available around here

Fair enough, but then wouldn't it make more sense to have a check/uncheck option in the actual SVP menu to enable/disable this?  For example, VLC has an option to enable/disable the tray icon.

88

(423 replies, posted in Using SVP)

Chainik wrote:

> there seems to be no way to have SVP running without the SVP

run it with "-tray" arg

I realize I'm pretty much literally a month late on this, but this is exactly what I was looking for, especially since it makes SVP behave like it does on Windows and the program even remembers this setting! (so when you have a video file associated with SVP, it only launches the SVP tray and not the SVP control panel window).

However, this does beg the question...why is this not the default behavior?  As I mentioned, this is how it behaves on Windows, so I would think that consistent behavior between different platforms would be ideal...

89

(423 replies, posted in Using SVP)

doemEINS wrote:

Anyone else facing completion errors while installing SVP4 on Ubuntu 16.04?

https://www2.pic-upload.de/thumb/33116398/Screenshotfrom2017-05-0422-42-54.png

Doesn't matter if I run setup as root or not, nor if I install it into home directory or / or /opt or else...


I just had the exact same issue on a relatively fresh installation of Linux Mint 18.1 Cinnamon, and running the following command solved it:

sudo apt-get install libqt5concurrent5 libqt5svg5

The thing is, I already manually installed Qt 5.50 before installing SVP, so I thought I didn't have to run that command...well I was wrong.  Run that command and everything should be good.


Unfortunately, I seem to still be having the issue where there seems to be no way to have SVP running without the SVP control panel also being open - clicking the X shuts down SVP completely, and associating a video file with SVP results in it opening both mpv and the SVP control panel.  If I have mpv automatically go to fullscreen, then yes it makes the SVP control panel not visible, but once the video ends it will only automatically close mpv and therefore leave the SVP control panel sitting there being open.



I did at least figure out why I didn't have a close button on the SVP Control Panel - for some reason, some of the themes results in there being no close button even though other non-default themes do have a close button:

90

(185 replies, posted in Using SVP)

Mystery wrote:

Decoding performance generally isn't an issue on computers

It's important in the case of Xvid/Mpeg4p2/Mpeg4ASP (and/or VP8 if the decode performance is similar) since the only semi-modern reason a person would use that is if they didn't have hardware decoding for h.264.


H.264 hardware decoding was only present on discrete GPUs starting around 2007 while Intel's iGPUs didn't include it until around 2008 - this means that many Core 2 Duo laptops are completely missing h.264 hardware decoding, but because the chipset supports AHCI they can still properly support SSDs and therefore long out-live their expected life.

Remember, the much newer low-power AMD Jaguar/Puma and Intel Silvermont architectures are still slower per GHz than a Core 2 Duo.

91

(185 replies, posted in Using SVP)

So I was doing a bunch of tests via Avidemux between the Xvid Mpeg4p2 encoder and the ff Mpeg4p2 encoder.

Not counting the likes of single and double-pass encoding, it seems like the settings that encode the quickest also perform the worst when playing back the video and vice-versa.


Basically, if you want the fastest encode, use the ff encoder with thread count set to "auto" (since it defaults to two) and quantization set to "mpeg".

Conversely, if you want the fastest decode, use the xvid encoder with quantization set to "h.263".



EDIT: Oh, and for whatever reason using an MP4 container results in slightly better decoding performance than an MKV container without compression.

92

(2 replies, posted in Using SVP)

One thing you can try is the old SVP 3.1.7 (which is free) - it's interpolation settings are very similar to those found in SVP 4 Pro, though it's not as user friendly.

Now there are several things in SVP 4 Pro that are missing in both SVP 4 Free and SVP 3.1.7, but you can at least use it to sample how the actual interpolation differs between Free and Pro.

Have you instead tried installing madVR via SVP's "Additional programs and features"?

94

(185 replies, posted in Using SVP)

Mystery wrote:

Were you using GPU acceleration? By default, it used GPU-acceleration if madVR was enabled in setting (I know, not a very obvious setting!) The new algorithm will unfortunately have no GPU acceleration, and it will be slower. I don't know yet to what extent. If you weren't using GPU acceleration already, then it won't make much difference.

Oh, yeah, GPU acceleration was not being used because for some reason it's actually slightly slower with it enabled.


Mystery wrote:

Intel Pentium G3258 3.2GHz has a CPU benchmark of 3928. Mine (from 4 years ago) is about 7600 -- about 2x faster.

1. I was running overclocked at 4.6GHz, which is 43.75% faster than stock clocks.

2. My CPU is only a 2core/2thread CPU, and this sort of task loves multiple threads (Ryzen would be a beast!).

3. Haswell, the architecture used in the Pentium G3258, launched 4 years ago.  If your CPU is also 4 years old and you're getting around 2x faster performance vs my stock clocks, then I'm going to guess you're using a Haswell quad core like the i5-4670.

95

(185 replies, posted in Using SVP)

Mystery wrote:

Wrong. You can change it yourself. Click on the Script tab and edit the Interframe parameter to the desired values.

Didn't realize that you can actually change the stuff there; I thought it was for copying and debugging and stuff.


Mystery wrote:

But then if I allow XVid and VP8, I need to allow additional container formats

Not true - Xvid in particular is supported in MP4.

VP8, yeah, but WebM is just a subset of MKV.  Also MKV works for Xvid too.


Mystery wrote:
Nintendo Maniac 64 wrote:

Another thing NGP doesn't seem to be able to do is slow down or speed up the audio

Same as above.

I've no idea how to do this via scripting...


Mystery wrote:
Nintendo Maniac 64 wrote:

Oh, and NGP's background colors don't work well with Windows' high-contrast themes that use white text.

hum...

Yeah, it's really quite hard to see the script text with the background image (see attached screenshot).


Mystery wrote:

The next version will have further changes that will considerably improve frame interpolation results.

Will it makes encoding slower?  Even with the fastest settings, I'm only seeing like 0.3x realtime encoding speed on a 1024x600 48.75fps video with 1.333x interpolation on my 4.6GHz Pentium G3258 (though presumably encoding to Xvid would make things faster).

96

(185 replies, posted in Using SVP)

Hey, so I just discovered a couple weeks ago that my first gen Intel Atom netbook actually has an IPS screen that looks way better than the TN screen on my AMD E-350 "not"book.  So for hour+ long videos, I've been using the netbook instead.



Problem is, the first gen Intel Atom is too slow to decode h.264 in software at the display's native resolution (1024x600) and the graphics is old enough (Intel 945GMA) that it lacks hardware h.264 decoding.

So I've found out that xvid actually works great on my netbook at 60fps with a resolution of 1024x600, so I already re-encode certain videos.

However, I noticed that NGP doesn't have a GUI option for xvid (though ffmpeg supports it), so because I don't know how to use ffmpeg commands, I have to re-encode the video a second time in xvid after NGP's own video encode.

It's worth noting that VP8 should also be decently lower in CPU utilization than h.264, but the program I use for encoding (Avidemux) doesn't include encoding support for it, so I've never tried a native 60fps 1024x600 video in VP8.  VP8 in particular would be useful for creating webms, though xvid would be useful simply for its fast encoding (relative to x264).



Another thing I discovered is that my netbook actually supports a refresh rate of 65Hz maximum with the stock EDID (I also did a bunch of tests and yes, it is not dropping or repeating any frames at 65Hz); this is important because custom resolutions are impossible in Windows with the Intel 945GMA, and its video playback performance in Linux is dreadful.

Doing some math, with "Uniform" interpolation, I discovered that 50fps videos slightly slowed down to 48.75fps would only need to display 3 interpolated frames for every source frame to hit 65fps - that would have considerably less artifacting than interpolating 50fps to 60fps (which requires 5 interpolated frames for every source frame).

...but NGP doesn't support custom frame rates, so I have to instead tell it to slow down the video by 90% to 45fps and then interpolate to 60fps, and then afterwards manually speed up the video playback speed to 65fps.



Another thing NGP doesn't seem to be able to do is slow down or speed up the audio like one could do in, say, Audacity by using the lossless "Set rate" function to a lower sampling rate (for 50fps with 48000Hz audio played back at 48.75Hz, you'd need the audio at exactly 46800Hz).  However, the audio part is very easy for me so it's not that big of a deal that it's missing.



So ideally, I would love you forever if NGP had GUI options for:

- xvid and/or vp8
- interpolating to a custom frame rate
- slowing down and/or speeding up the audio (with or without optional pitch correction)


Oh, and NGP's background colors don't work well with Windows' high-contrast themes that use white text.


Tkompuras wrote:

Hello guys, I am a newbie with this so bear with me. So I want to convert a mkv file to 60fps using yin media encoder but I get an error (On the command line it vanishes so fast and I can't see it) when I hit preview changes or play in MPC-HC, the error states something related to ConvertToShader. Can anyone help?

I had this problem myself with v1.4, but trying v1.32 instead worked perfectly fine, so try that (the media encoder is very similar between the two).

Interestingly enough however, when I reinstalled and tried v1.4 again just today, it worked just fine...I wonder if v1.32 installs something that v1.4 needs?

97

(13 replies, posted in Using SVP)

Chainik wrote:

nope

You're right, I just did a test that confirmed it.

...that's annoying to say the least.

98

(13 replies, posted in Using SVP)

Wait - doesn't SVP process and output at 4:4:4?

Therefore, shouldn't you get the same 4:4:4 chroma sub-sampling regardless of whether it's SVP or the video player that's downscaling>

99

(5 replies, posted in Using SVP)

Chainik wrote:

Ensure that you do your tests with "frame interpolation mode" = "uniform"

Oh so that's why Uniform takes an absurdly more amount of computational grunt when doing 50fps --to-> 60Hz.

But for my offline rendering tests, AFAICT, I cannot set uniform interpolation since the program I'm using seems to have some sort of preset settings for SVP interpolation.

100

(5 replies, posted in Using SVP)

dejavecue wrote:

I agree that it must be smoother, but interpolating to 300 fps from 50 fps also means that you add a lot of "fictional" frames and thus drastically lose information (relatively speaking), so that image quality would suffer.

While this is true, it's worth noting that this increase in artifacting is much less with something like 60fps --3x-> 180fps --1/2x-> 90fps; in fact, the artifacting would be less than plain old 30fps --3x-> 90fps.