Vapoursynth filter doesn't load property page about avisynth_filter

[CrendKing]

It looks like a registration issue. Try regsvr32 vapoursynth_filter_64.ax. It should give you an error. One possible cause is that you didn't have VapourSynth directory in your PATH, more specificallyVSScript.dll.

[Usulyre]

I tried regsvr32 vapoursynth_filter_64.ax and it succeeded with no error. How do I add VapourSynth directory in your PATH, more specifically VSScript.dll?

[CrendKing]

How do I add VapourSynth directory in your PATH, more specifically VSScript.dll?

If you used VapourSynth installer, that option should already be checked by default. How did you install VapourSynth? Portable?

[Usulyre]

Installed the x64 installer version, i believe the options was checked but i can try to install again

[Usulyre]

what options should i choose for python and vapoursynth when installing?

[CrendKing]

Python: First page, "Add Python 3.9 to PATH"
VapourSynth: "Add VSPipe and AVFS to PATH"

[Usulyre]

I got it to work now. I didn't see "Add Python 3.9 to PATH" but I did see something like "add python to environmental path" or something like that. I believe I already checked that before. I did check a few options more for python - pip and py launcher and associate files with python. For vapoursynth i checked also vsrepo package manager and add new vapousynth python script to shell context men and update vsrepo package list and add vsrepo to path as well.

Going to try that filter i mentioned above. Have you tried it yet, does it work with vapoursynth filter?

[Usulyre]

shell context menu

[Usulyre]

Not sure but how does one set a target framerate with this filter?

Here is my script,

Imports

import vapoursynth as vs
core = vs.get_core()

clip = VpsFilterSource
clip = core.resize.Bicubic(clip=clip, format=vs.RGBS, matrix_in_s="470bg", range_s="full")

clip = core.rife.RIFE(clip, model=0, gpu_thread=4, tta=False, uhd=False, sc=True, fp32=False, list_gpu=False)

[CrendKing]

First, according to the project, "Only planar RGB format with float sample type of 32 bit depth is supported." DirectShow does not support outputting planar RGB format, so we need to convert it back to YUV.

Second, your script is using sc=True, but according to the project, "You must invoke misc.SCDetect on YUV or Gray format of the input beforehand so as to set frame properties."

Third, according to the parent project, this thing "Input two frame images, output one interpolated frame image." So it can only double the fps.

---

Here is my script to get correct frames. However, the output rate is extremely low, making the video choppy. For example, with a 30fps video, normally AVSF deliver each frame per 33ms. Adding the RIFE() call increases that to 92ms. The renderer is forced to drop lots of frames. But my CPU and GPU are all pretty much idle. Maybe this filter is not designed for realtime use?

My script:

    sc_clip = vapoursynth.core.misc.SCDetect(VpsFilterSource)
    rgb_clip = vapoursynth.core.resize.Bicubic(sc_clip, format=vapoursynth.RGBS, matrix_in_s="470bg", range_s="full")
    rife_clip = vapoursynth.core.rife.RIFE(rgb_clip, sc=True, fp32=True)
    output_clip = vapoursynth.core.resize.Bicubic(rife_clip, format=vapoursynth.YUV420P8, matrix_s="709")
    output_clip.set_output()

[Usulyre]

Right I see that it is only doubling the original framerate.

Although I cannot see any difference visibly in the video, just noticed the framerate going up.

I couldn't get your script to work for some reason.

I will ask if there is a way to increase the framerate more.

My updated script

import vapoursynth as vs
core = vs.get_core()

clip = VpsFilterSource
clip = core.resize.Bicubic(clip=clip, format=vs.RGBS, matrix_in_s="470bg", range_s="full")
clip = core.rife.RIFE(clip, model=1, gpu_id=0, list_gpu=True, gpu_thread=4, tta=True, uhd=False, sc=False, fp32=True)
output_clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P8, matrix_s="709")
output_clip.set_output()

[Usulyre]

I also noticed this:

https://github.com/HomeOfVapourSynthEvolution/VapourSynth-RIFE-ncnn-Vulkan/issues/2

I don't know myself but is that script setting a custom framerate?

[CrendKing]

I'm pretty sure if you use list_gpu=True, it won't do any interpolation. You can verify the theoretical output FPS in AVSF's Status page.

[Usulyre]

In the status page with my script unchanged the output fps is the same as input, but in mpc-be statistics page, it is doubled. When I took out list_gpu=True, video playback was choppy and slow and the framerate was like 3 fps?

[CrendKing]

You can see the source code that if list_gpu=True, it just invoke Text and return. It doesn't load model and do anything real processing below.

[CrendKing]

So if it turns out there is no issue in AVSF, I'm closing this.