Vapoursynth

[Selur]

@lansing:
The point was if anyone did have a function which would take:
a. a clip
b. rectangle coordinates+size
c. a filter + parameters
d. a frame range
which would:
only apply the filter on the rectangle and the frames.

So the function basically should only for the frames in the range.
1. take the source, crop out the recangle
2. filter the rectangle
3. replace the space of the rectangle with the filtered rectangle
The main point is not filter the whole clip and all the whole frame.

@_AI_: I somehow totally missed the link to the animate.py, that does help.
Thanks! I need to play around with this. :)

Cu Selur

[Myrsloik]

R57 (https://github.com/vapoursynth/vapoursynth/releases/tag/R57)!

r57:
added close method for frames in python (cid-chan)
arguments names will no longer have leading _ stripped in python, append _ to the end or argument names instead to avoid python keywords
fixed nodes never being marked as nfnocache/nfiscache in api3 which caused some older applications to enter an infinite cache insertion loop
fixed memory bloat in python due to circular references when dealing with frame objects
fixed vsrepo missing genstubs files
fixed vfw and avfs speed regression from r55
fixed mismatched format clips not working regression from r55

[ChaosKing]

Vapoursynth still has no official logo. How about something like this?

https://i.imgur.com/Kp7yTaB.png
https://i.imgur.com/VCyqIVE.png
https://i.imgur.com/UIZ4bpG.png
https://i.imgur.com/aLfzZ2L.png

[LigH]

Once upon a time I suggested something based on the Python snake icon and a film roll.

https://www.ligh.de/pics/VapourSynth_FilmClip.png

[lansing]

Last month was Vapoursynth 9th anniversary too

[poisondeathray]

I liked LigH's idea, because python is a central component to vapoursynth, and animated the filmstrip
https://forum.doom9.org/showthread.php?p=1847472#post1847472

Here is a revised version with text
https://i.postimg.cc/Vkv1q49J/vapoursynth-pythonfilm-v3-320x320.gif

[l33tmeatwad]

You can't just use the Python logo without permission as it's copyrighted.

[LigH]

Thank you for the reminder; still, one could ask for permission when it gets serious...

[lansing]

How do I load in an audio file in the script? I couldn't find it in the documentation

[l33tmeatwad]

Use best audio source, then you'll have to set two outputs, one for video and one for audio.

[lansing]

Use best audio source, then you'll have to set two outputs, one for video and one for audio.

I tried to load this in media player classic but it gives me the error "output index 0 is not video"


clip = core.dgdecodenv.DGSource(r"clip.dgi")
audio = core.bas.Source(r"audio_file.flac")
clip.set_output()
audio.set_output()

[l33tmeatwad]

You need to put 0 for video and 1 for audio in the parenthesis.

[lansing]

You need to put 0 for video and 1 for audio in the parenthesis.

Now the program crashed with "access violation"

"crashing modules: C:\Program Files\VapourSynth\core\vsvfw.dll"

[Izuchi]

How do I output 32-bit float video using vspipe? It seems to just give me green garbage.

[Yomiko]

How do I output 32-bit float video using vspipe? It seems to just give me green garbage.

That looks right if the output is interpreted as YUV in integer format

[poisondeathray]

How do I output 32-bit float video using vspipe? It seems to just give me green garbage.



YUV444PS or RGBS ? Not many programs will understand YUV444PS

What program are you piping into ?

For ffmpeg/libavcodec, you can vspipe rawvideo as RGBS and use -pix_fmt gbrpf32le in ffmpeg

[Izuchi]

Was hoping to output YUV420PS as an intermediary file after running BM3D CUDA and then inserting it back into the script. Is RGB my only option?

[poisondeathray]

Was hoping to output YUV420PS as an intermediary file after running BM3D CUDA and then inserting it back into the script. Is RGB my only option?


Even if you vspipe it to something , I do not know of any file format that you could read for YUV420PS. The raw reader in vapousynth does not support float, only up to 16bit int formats. All vapoursynth imwri reads are in RGB , even if source was YUV. (jpeg, YUV tiff)

For intermediate files, you could stay in vapoursynth by using imrwi to read/write a RGBS TIFF sequence, without vspipe or other programs . (The current imwri.write implementation only writes RGBH in EXR, but RGBS is ok in TIFF). If you use resize.Point it should be lossless from YUV420PS <=> RGBS . Make sure you use float_output=True for the imwri.Read step.


And how did you YUV420PS to work in vapoursynth ? I couldn't import it from core , and get "module 'vapoursynth' has no attribute 'YUV420PS'" error

[Izuchi]

Yeah YUV420PS didn't work for me either, it just returned garbage pixels ffms2 just reads it as YUV420P8 with only a single frame. Thanks for tip, I might try piping it to ffmpeg.

[Izuchi]

Would this be the correct command? vspipe script.vpy - | ffmpeg -f rawvideo -pix_fmt gbrpf32le -s 1920x1080 -r 24000/1001 -i - denoise.rgb

[_Al_]

no attribute 'YUV420PS'
former register_format(), now it is query_video_format(), I guess because of audio formats addition:
my_float_format = core.query_video_format(color_family=vs.YUV, sample_type=vs.FLOAT, bits_per_sample=32, subsampling_w=1, subsampling_h=1)
clip2 = clip.resize.Point(format=my_float_format)
print(clip2)
>>>
VideoNode
Format: YUV420PS32

[poisondeathray]

Would this be the correct command? vspipe script.vpy - | ffmpeg -f rawvideo -pix_fmt gbrpf32le -s 1920x1080 -r 24000/1001 -i - denoise.rgb


For raw RGB... but you can't read raw rgb float with vsrawsource. You would use denoise_%05d.exr for ffmpeg output, and imwri.Read with float_output=True for import back into vapoursynth. But there are issues with that workflow for some reason, it's not lossless if you check MakeDiff, I'm looking into it.

But if you use imwri.Write in vapoursynth it's lossless

output

clip = core.imwri.Write(clip, "TIFF", "imwri_Write_%05d.tiff",firstnum=0)


reimport

clipc = core.imwri.Read(r'imwri_Write_%05d.tiff', float_output=True)

[poisondeathray]

former register_format(), now it is query_video_format(), I guess because of audio formats addition:
my_float_format = core.query_video_format(color_family=vs.YUV, sample_type=vs.FLOAT, bits_per_sample=32, subsampling_w=1, subsampling_h=1)
clip2 = clip.resize.Point(format=my_float_format)
print(clip2)
>>>
VideoNode
Format: YUV420PS32

Thanks it works.

I remember asking about not included pixel formats and Holy Wu told me about register_format - I should have remembered

https://forum.doom9.org/showthread.php?p=1904930#post1904930

[Myrsloik]

Thanks it works.

I remember asking about not included pixel formats and Holy Wu told me about register_format - I should have remembered

https://forum.doom9.org/showthread.php?p=1904930#post1904930

How did you end up with YUV420PS in the first place? Subsampled float is a quite weird thing to have

[poisondeathray]

How did you end up with YUV420PS in the first place? Subsampled float is a quite weird thing to have

It was not for me, scroll up a few posts

[Quadratic]

How did you end up with YUV420PS in the first place? Subsampled float is a quite weird thing to have

Why is YUV420PS "weird"? It can be achieved with vsutil and fmtc. It is not uncommon.

[Myrsloik]

Why is YUV420PS "weird"? It can be achieved with vsutil and fmtc. It is not uncommon.

It's a very inefficient use of bits. You'd have an image that definitely looks better with full resolution chroma in YUV444P16 format and smaller size.

[Julek]

Now I need some input on which frame statistics to implement. If you generalize all of the functions in avisynth you get these 6 left. I've seen that difference is used a lot but are there any of these I could skip? I can't remember seeing the min/max/median used ever.
You can also suggest new ones if they're not too complicated.

Average(clip)
Difference(clip1, clip2)
PlaneMax(clip, float threshold)
PlaneMin(clip, float threshold)
PlaneMedian(clip)
PlaneMinMaxDifference(clip, float threshold)

The full list here:
http://avisynth.org/mediawiki/ScriptClip
Any chance to implement threshold for Min/Max now?
Dogway's retinex uses this, I know there is a plugin for it, but the output is not the same and I wanted to do some tests, if possible.

[Dogway]

Vapoursynth still has no official logo. How about something like this?

https://i.imgur.com/Kp7yTaB.png
https://i.imgur.com/VCyqIVE.png
https://i.imgur.com/UIZ4bpG.png
https://i.imgur.com/aLfzZ2L.png

Beautiful logos! I like 3rd and 4th the best. 4th looks more like consumer oriented, it's third or fourth in neutral tones.

[groucho86]

Hi everyone,
I opened an issue on github (https://github.com/vapoursynth/vapoursynth/issues/823) but thought it may be worth posting here. I installed Vapoursynth R57 and imwri via homebrew and am trying to load a half-float EXR (https://github.com/vapoursynth/vapoursynth/files/7398013/halffloat.001.exr.zip).

clip = core.imwri.Read(filename='halffloat.%03d.exr', firstnum=1,
float_output=True)
print(clip.get_frame(0))

Returns:
Filter Read returned a frame that's not of the declared format
VapourSynth encountered a fatal error: Filter Read returned a frame that's not of the declared format
libc++abi.dylib: terminating

float_output set to False works.

[Yomiko]

What's the reason for not allowing a GRAY plane extracted from an RGB clip resized to RGB because of its tagged matrix? It seems the matrix itself (if supported by VS) doesn't affect the output result.

[vxzms]

Would you plan to let the vspipe’s filter-time also show namespace? I think it would be more intuitive.

[Myrsloik]

Would you plan to let the vspipe’s filter-time also show namespace? I think it would be more intuitive.

Filter don't have a namespace but the functions that create them may or may not. So it's actually very hard to add what you'd expect as the namespace to your list.

If anything extending the graph output with time spent in various filters is probably the best way since then you can see which function is responsible.

[GB452]

Having a very strange issue: vsrepo can't detect VapourSynth. I had updated to VS 57, but had some issues (including this one) and uninstalled it and went back to what I was using previously, version 54, but I keep getting this error and I cannot find any suggestions on how to fix it online.

https://i.imgur.com/yFFP3lE.png

[ChaosKing]

Same happend to me after updating to R57.

[Yomiko]

What's the reason for not allowing a GRAY plane extracted from an RGB clip resized to RGB because of its tagged matrix? It seems the matrix itself (if supported by VS) doesn't affect the output result.

Meanwhile GRAY clip with unspec matrix can be resampled. A bug?

[Myrsloik]

Meanwhile GRAY clip with unspec matrix can be resampled. A bug?

No, gray clips are the Y in YUV so the same rules apply. You can resize unknown matrix YUV but you can't convert it to RGB.

[Selur]

I'm on Ubuntu 20.04, I compiled and installed Vapoursynth R57.
I also compiled a bunch of filters to "/home/selur/opt/vapoursynth/lib/python3.8/site-packages"

paths seem to be correct:

echo $LD_LIBRARY_PATH
/home/selur/opt/vapoursynth/lib
echo $PYTHONPATH
/home/selur/opt/vapoursynth/lib/python3.8/site-packages
echo $PATH
/home/selur/.local/bin:/home/selur/opt/vapoursynth/bin:/home/selur/.local/bin:/home/selur/opt/vapoursynth/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin



import vapoursynth as vs
core = vs.core
clip = core.std.BlankClip(format=vs.RGB24, color=[255, 255, 255])
clip = core.text.Text(clip, core.version())
clip.set_output()

works fine

explicit loading a library (libvslsmashsource.so):

import vapoursynth as vs
core = vs.core

core.std.LoadPlugin('/home/selur/opt/vapoursynth/lib/vapoursynth/libvslsmashsource.so')

clip = core.lsmas.LWLibavSource(source="test.mov", format="YUV422P10", cache=0, fpsnum=30000, fpsden=1001, prefer_hw=0)
clip = core.resize.Point(clip, matrix_in_s="470bg",range_s="limited")

clip.set_output()
works fine too,

but using:

import sys
import vapoursynth as vs
core = vs.core

sys.path.append('/home/selur/opt/vapoursynth/lib/vapoursynth')

clip = core.lsmas.LWLibavSource(source="test.mov", format="YUV422P10", cache=0, fpsnum=30000, fpsden=1001, prefer_hw=0)
clip = core.resize.Point(clip, matrix_in_s="470bg",range_s="limited")

clip.set_output()

doesn't.
It returns "Python exception: No attribute with the name lsmas exists. Did you mistype a plugin namespace?"

-> How can I tell Vapoursynth where to look for the libraries?
I though:
sys.path.append('/home/selur/opt/vapoursynth/lib/vapoursynth')
would do the trick, which it doesn't.

Cu Selur

[Myrsloik]

You're specifying the python search path, it's only used for import and similar python commands. You also shouldn't install filters in site-packages.

[Selur]

Sorry, my mistake. Didn't put the iflters in the site-packages. :)
site-packages only contains vapoursynth.la and vapoursynth.so
all compiled filters are in "/home/selur/opt/vapoursynth/lib/vapoursynth/"

Okay, so is there a way to somehow to use my folder as "/home/selur/opt/vapoursynth/lib/vapoursynth/" source for autoloading?

[Selur]

okay, forget it found it.
"$HOME/.config/vapoursynth/vapoursynth.conf "
with:
"SystemPluginDir=/home/selur/opt/vapoursynth/lib/vapoursynth"
works

[Yomiko]

Also this -> http://www.vapoursynth.com/doc/functions/general/loadallplugins.html

[Selur]

Uhhh,.. didn't know "LoadAllPlugins" :)

[_Al_]

oh, was LoadAllPlugins introduced with new API release or does it work with older releases too?

[lansing]

When creating filter, where do we put the xyvInit() initial function in API4 now? It was gone in the example filter. And in the createVideoFilter(), there are this two lines in the example script:


VSFilterDependency deps[] = {{d.node, rpStrictSpatial}};
vsapi->createVideoFilter(out, "Invert", vi, invertGetFrame, invertFree, fmParallel, deps, 1, data, core);


What is the VSFilterDependency and numDeps do? The documentation is lacking.

[Myrsloik]

When creating filter, where do we put the xyvInit() initial function in API4 now? It was gone in the example filter. And in the createVideoFilter(), there are this two lines in the example script:


VSFilterDependency deps[] = {{d.node, rpStrictSpatial}};
vsapi->createVideoFilter(out, "Invert", vi, invertGetFrame, invertFree, fmParallel, deps, 1, data, core);


What is the VSFilterDependency and numDeps do? The documentation is lacking.

1. You have a function that calls createVideoFilter(), put it there instead. Most filters are already written that way.

2. It's an array where you list all the nodes and in which order you will request frames from them to generate the output. StrictSpatial is obviously for spatial only filters, NoFrameReuse is for reordering filters of certain types (basically if you request all output frames once none of the input frames will be requested twice, trim, splice and similar fit into this category). General is for everything else.

NumDeps is simply how items are in the deps array.

[lansing]

Okay got it. Also can you add some example filter written in c++? Because when I got stuck with the c syntax, I couldn't find any other example because everybody was writing in c++. For example, I want to use the isConstantVideoFormat() from the VSHelper4.h, I included the header file in my c file #include "VSHelper4.h", but visual studio still couldn't recognize the function.

[DJATOM]

Use vsh::yourCalledFunction.

[lansing]

Use vsh::yourCalledFunction.

using namespace is a c++ thing

[WolframRhodium]

vsh_yourCalledFunction