Search options

I was reading the documentation for NVOFA FRUC (see https://docs.nvidia.com/video-technolog … index.html), and it looks like this would make optical flow-based frame interpolation possible for any NVIDIA GPUs since the Turing generation.

I suppose there needs to be 2 different options offered now. 1) Only use motion vectors from NVOFA and let SVP do the rest, and 2) Allow NVOFA FRUC to generate the interpolated frames.

(2) is new since SDK 4.0 and would not have been easily possible before. I don't know how the FRUC frames' quality compares with SVP, but it would be an appreciated feature regardless of if it is technically worse. There is sample code provided from NVIDIA that shows how relatively simple it is to integrate NVOFA FRUC into applications.

I just had a question about using SVPflow directly on Linux (with vspipe and ffmpeg). If my source is 10-bit HDR and I want to preserve this in the raw video output from SVPflow, is there a way to do this, or will it always convert it to YV12?

I ask because I was reading through the SVPflow manual and saw this near the bottom of the page under "Basic Vapoursynth script":
clip = # need some input here
clip = clip.resize.Bicubic(format=vs.YUV420P8) #convert to YV12

Can this format be YUV420P10 instead or will that mess up SVP's calculations? Can HDR information be preserved?

If not, I'm fine with encoding SVP's raw video output to 8-bit hevc using ffmpeg (rather than 10-bit hevc) so long as SVPflow converts the 10-bit content to YUV420P8 faithfully. Thanks for any info!

Hi again,

Since I'm using vapoursynth scripts on Linux, I don't need SVPManager but I DO need updated svpflow plugins. The ones on the Download page are very outdated (4.2.0.xxx vs 4.3.0.xxx).

Do I have have to manually extract them from the ".run" binary every new release or am I just missing where new ones are hosted?

Thanks!

My vapoursynth/SVP script: https://pastebin.com/MC9EFFUG

My laptop's memory is about 8GB and within 4-5 minutes of running this script (piped to ffmpeg which is fine, not even 200 MB of ram), my computer freezes (I'm on Linux). I tried setting core.max_cache_size = 5000 and this did nothing.

The video I'm using is very large... 1080p Blu-ray rip, 24 GB. Perhaps the video size is too big? Or that I'm using ffms2 to read the video?

So, 2 questions. 1) How can I stop the script from maxing out my memory, and 2) How can I pass a video into this script without using ffms2? I know it must be blatantly simple but this is a whole new world for me :)

Thank you for ANY help!

P.S. Completely random but what does rc:true actually mean? The documentation says it's for SVP Manager, but in what way?

I've been experimenting with SVPCode (on Linux). I'd really like to just use ffmpeg and vapoursynth because customizing the encoding process (and everything really) is so much simpler without having to work around mpv.

Is it possible to use SVP (with GPU) by just doing "python foo.vpy | ffmpeg -i etc. -o foo_60.mkv" ? (obviously pseudocode, just as an example)

If it is possible, what raw video format does SVP output?

If it isn't possible, how can I pass these ffmpeg options (-hwaccel vaapi -hwaccel_device /dev/dri/renderD128 -hwaccel_output_format vaapi) to mpv so I can benefit from vaapi encoding/decoding?

I'd rather stick to the command-line rather than the GUI for SVPCode... just personal preference.

Thanks for any help!

I know video encoding is a subjective, messy topic when talking about "optimal" quality, but I am trying to understand this from the SVP point of view.

I'm getting a new laptop soon with an i5-8365U CPU and Intel UHD 620 iGPU. I will install either Gentoo or Arch Linux on it... Shouldn't matter as they both support the new "Intel-Neo" OpenCL and "Intel-Media-Driver" for VAAPI.

My goal will be to use SVPcode to re-encode somewhat lossy x264/x265 1080p content (not going to lie, acquired through "online sources"...) to H.265. I'd like to use FFmpeg's hevc_vaapi encoder to do so since it is way easier to set up than QSV (and still open-source).

Obviously, I want to avoid quality loss. Inserting interpolated frames adds to a file's data so I expect the resulting file to be larger than my source. How can I ensure the re-encoded file is minimally affected by the hevc_vaapi encoder? I assume setting "hevc_vaapi -q:v 18" and the other necessary options is good enough for no noticeable quality loss... Am I missing something?

Any help is greatly appreciated! :)