Hybrid: Input -> x264/x265/Xvid/VP8/VP9

[Selur]

Independent of the target audio format when reencoding.
If possible, bit depth is kept and sample rate is handled as set under Audio->Base->Audio Encoding Settings->Filtering.
By default Hybrid will resample to 48kHz, but you can enable 'Keep input sample rate' (if possible this will be kept).

[juanitogan]

Bug report for
Hybrid v2024.03.18.1:
Hybrid v2024.04.06.1:

Avisynth related:

Avisynth > Misc > Preview > [x] Show Frame number Positions: top right
This adds frame numbers to both preview and encoding (unlike what the hover help says).

Avisynth Preview window always closes and reopens when using the Refresh Preview button, losing frame number, window position and size.
Turnaround can be near a minute, depending on filters in use. The main Preview button does the same thing.
In case it helps, the log is filled with these messages from each use of the Preview:
> Couldn't find 'DePulse64' in the filter order, adding 'KillPulse32' at the end.
> using 32bit AvisynthViewer: C:\Program Files\Hybrid\32bit\avsViewer.exe

Avisynth Preview window does not remember window position and size.

Avisynth Preview window does not remember setting: [x] Scrolling. (Which can take several seconds to enable depending on filters.)

AutoAdjust > Gamma mode > hover shows TV output help, not Gamma mode help
AutoAdjust > Savety (misspelled Safety)

AutoAdjust (and similar auto tools) should probably go much further down the filter order by default. I can't imagine ever wanting to auto level before removing noise and artifacts.

Avisynth > DeNoise > QTGMC > enabling this breaks the ability to preview and encode (frame 0 only, stops responding, etc).

Avisynth > DeNoise > I can't get several filters to appear in the script: FFT3DFilter, FFT3DGPU, KNLMeansCL; but turning on mClean loads 8 dlls including FFT stuff.


Vapoursynth related:

(De-)Interlace/Telecine > VIVTC (Vapoursynth) > Misc > Fix combed with deinterlacer > Deinterlacer: QTGMC(preset="fast") > Preview
> Error: VsFixCombed - Unknown Deinterlacer: QTGMC(preset="fast")
Actually, none of the Deinterlace settings works here; not just QTGMC fast.

(De-)Interlace/Telecine > VIVTC (Vapoursynth) > VFM
Errors only with telecined source, I believe:
> File "C:\Program Files\Hybrid\64bit\vsscripts\havsfunc.py", line 2271, in QTGMC_Interpolate
> eedi3 = partial(core.eedi3m.EEDI3CL, field=field, planes=planes, mdis=EdiMaxD, device=device, **eedi3_args)
> ^^^^^^^^^^^^^^^^^^^
> File "src\\cython\\vapoursynth.pyx", line 2691, in vapoursynth.Plugin.__getattr__
> AttributeError: There is no function named EEDI3CL
Changing Deinterlacer to something besides one of the QTGMC options works. Same issue with TIVTC

(De-)Interlace/Telecine > VIVTC (Vapoursynth) > VDecimate
The Block (X/Y) fields have no effect. The script always shows a value of 16:
> clip = core.vivtc.VDecimate(clip=clip, blockx=16, blocky=16)# new fps: 20.2336

Vapoursynth > Misc > Filter Order/Queue
Turn on QTGMC but mClean gets bolded instead.

(De-)Interlace/Telecine > VIVTC (Vapoursynth) > VFM
Errors only with telecined source, I believe:
> File "C:\Program Files\Hybrid\64bit\vsscripts\havsfunc.py", line 2271, in QTGMC_Interpolate
> eedi3 = partial(core.eedi3m.EEDI3CL, field=field, planes=planes, mdis=EdiMaxD, device=device, **eedi3_args)
> ^^^^^^^^^^^^^^^^^^^
> File "src\\cython\\vapoursynth.pyx", line 2691, in vapoursynth.Plugin.__getattr__
> AttributeError: There is no function named EEDI3CL
Changing Deinterlacer to something besides one of the QTGMC options works. Same issue with TIVTC
[[ UPDATE: Still seeing this after loading a progressive and turning on QTGMC in the VS filters tab (after doing a telecine job).
In the deint tab, Auto-deint is set to None and [x] Overwrite... is greyed out (both checkbox and dropdown) and set to "telecine".
Fiddle with the deint settings until it shows [x] Overwrite... greyed out and set to "progressive" and it works again; very slow now.
And why is Auto-deint changing to match each new title load? I thought it usually just stayed at GTGMC. ]]
--> Restarting Hybrid now has Auto-deint back to QTGMC, even on progressive material, and [_] Overwrite... enabled but the dropdown disabled and set to "progressive".
This looks more normal and turning on the QTGMC filter in the VS tab now works fine.


Other:

Filter Config Save saves/sets individual Deinterlace/Telecine settings... but not which Auto Deint Handling was selected.
Very weird contradiction. Don't know how to save this per job... and would prefer it saved/set with Filter Config.

More of a feature add, I'm guessing: VIVTC/TIVTC/etc input and output frame counts are not correct for VFR sources (and could add a hint to manually set Change Speed [and/or fix that too]). Don't see a VFRtoCFR filter (or `ffmpeg -vsynch cfr` equivalent). Somehow this works without a CFR filter; not sure how. Oh, wait, I see AssumeFPS() in the script... but, no, that's not it since it contains the average VFR value. Anyway, I'm running into some confusion here that may or may not need work. Decimating telecined and interlaced VFR material _usually_ works well when I get the settings right... but it feels a bit like magic.

FDKAAC progress bar/text does not become visible until more than half done. This 20-second delay makes me poke around to make sure the job is running.

Hybrid writes APPDATA\hybrid\fontconfig\... folders in my working folder.

Audio cut short during detelecine (from a source flagged as telecine -- not an issue when decimating a source flagged as TFF). When setting audio to "custom" and encoding AC-3 to AAC FDK (no other codec tested), it cuts the audio short as if the length was changed by the number of frames the detelecine drops -- that is, it cuts off the last 1/5 of the audio track (presumably because it calcs the frames at 30fps still instead of 24fps). Tested "auto add (all)" and results were the same (but file size was 3kb [out of 50mb] larger somehow). Tested "auto add (all)" with multiple input files and IT WORKS!... very weird workaround, if you can call it that (I stumbled upon this by accident). I checked the temp *.aac file and it is also short, so not a muxing problem.

The completion bar for the video during detelecine (including decimating while deinterlacing) hits 100% 4/5 of they way through and keeps going to 125%. Not sure if this is intentional or not. I can't see the audio bar due to the other issue that hides it for the first 20 seconds or so, at which point it is already done with this material.

1920x1080 PAR 0.75 stretches Preview wrong way to 2560x1080 instead of 1440x1080 (and, yes, it was a weird to find such a source). I didn't try encoding it to see what happens. Instead, I forced a resize to 1440x1080 with 1:1 pixels before encode.

Bookmarks are lost whenever I reset the VS Preview. I would prefer it renamed the last tempPreviewVapoursynthFile*.vpy.bookmarks file to match the new tempPreviewVapoursynthFile*.vpy on refresh to keep the bookmarks.

Also, I saw you saw the report I made on DeScratch's memory leak and the new version released. Just a reminder here to put it in (it needs a new version of VS [API 4.1 minimum] and Python, etc.). Update: I see you talked him into rebuilding for API 4.0. Awesome!


I like the new Artifacts tab!


Suggestions:

- Add a field to the VS preview that shows the current frame's time.

- Add similar TV/PC presets to Vapoursynth > Color > SmoothLevels that I see on AutoAdjust.
That, or at least remind us in the hover help to set Input or Output to 16/235 for TV.

- Put a yes/no warning message box on Defaults > Reset All.
I've made the mistake more than once now (when I get tired, I guess) thinking it will reset to my last save. Ack!
Maybe add another option to reset to last save (same as a Restart, I suppose).

- Under custom, add an optional interface for setting bits and/or colorspace (that adds the appropriate line), so Hybrid can track it and doesn't have to do unnecessary conversions after it. Yeah, that's a big one, I'm sure.

- Don't add a Start chapter if there is a chapter less than 1 second from start. MakeMKV rarely (ever?) creates a chapter at 0.000, it seems. (Could maybe even reduce that to less than 0.1 seconds, if you want, from what I've seen so far from MakeMKV.)


Thanks again for this excellent tool!

Matt

[Selur]

Avisynth related:

Avisynth > Misc > Preview > [x] Show Frame number Positions: top right
This adds frame numbers to both preview and encoding (unlike what the hover help says).

hopefully fixed now

Avisynth Preview window always closes and reopens when using the Refresh Preview button, losing frame number, window position and size.

Can't reproduce here.

Turnaround can be near a minute, depending on filters in use.

Yes, depending on the filters preview is slow.

The main Preview button does the same thing.

yes, that is as expected.

In case it helps, the log is filled with these messages from each use of the Preview:
> Couldn't find 'DePulse64' in the filter order, adding 'KillPulse32' at the end.
> using 32bit AvisynthViewer: C:\Program Files\Hybrid\32bit\avsViewer.exe

My guess is that you are running out of RAM that Avisynth can use.
Try lowering the thread count and if that does not help switch to 64bit.

The 'DePulse64' and 'KillPulse32' info should not cause problems.
This probably happens since you haven't: properly uninstalled Hybrid with settings, migrated your settings, reset your defaults when it was mentioned in the changelog

Avisynth Preview window does not remember window position and size.

Avisynth Preview window does not remember setting: [x] Scrolling. (Which can take several seconds to enable depending on filters.)[/quote]
Correct. It never did. Patches to avsViewer (https://github.com/Selur/avsViewer) are welcome.

AutoAdjust > Gamma mode > hover shows TV output help, not Gamma mode help

hopefully fixed now, also needs to be renamed, it's not 'Gamma mode' but 'Gain mode'

AutoAdjust > Savety (misspelled Safety)

hopefully fixed now

AutoAdjust (and similar auto tools) should probably go much further down the filter order by default. I can't imagine ever wanting to auto level before removing noise and artifacts.

I'm not planning to change the default order.

Avisynth > DeNoise > QTGMC > enabling this breaks the ability to preview and encode (frame 0 only, stops responding, etc).

Works fine here, I suspect you are running out of RAM

Avisynth > DeNoise > I can't get several filters to appear in the script: FFT3DFilter, FFT3DGPU, KNLMeansCL; but turning on mClean loads 8 dlls including FFT stuff.

hopefully fixed now

Vapoursynth related:

(De-)Interlace/Telecine > VIVTC (Vapoursynth) > Misc > Fix combed with deinterlacer > Deinterlacer: QTGMC(preset="fast") > Preview
> Error: VsFixCombed - Unknown Deinterlacer: QTGMC(preset="fast")
Actually, none of the Deinterlace settings works here; not just QTGMC fast.

Works fine here, fixed it a while ago.

(De-)Interlace/Telecine > VIVTC (Vapoursynth) > VFM
Errors only with telecined source, I believe:
> File "C:\Program Files\Hybrid\64bit\vsscripts\havsfunc.py", line 2271, in QTGMC_Interpolate
> eedi3 = partial(core.eedi3m.EEDI3CL, field=field, planes=planes, mdis=EdiMaxD, device=device, **eedi3_args)
> ^^^^^^^^^^^^^^^^^^^
> File "src\\cython\\vapoursynth.pyx", line 2691, in vapoursynth.Plugin.__getattr__
> AttributeError: There is no function named EEDI3CL
Changing Deinterlacer to something besides one of the QTGMC options works. Same issue with TIVTC

Yeah, that's a error I know about, log should probably show something like:

Not loading, 'EEDI3m_opencl' for vsFixCombed since 'EEDI3m' is already loaded.

Wrote a workaround, but not sure it catches all cases. (this will leat to that whatever gets first loaded EEDI3m or EEDI3CL get always used)

(De-)Interlace/Telecine > VIVTC (Vapoursynth) > VDecimate
The Block (X/Y) fields have no effect. The script always shows a value of 16:
> clip = core.vivtc.VDecimate(clip=clip, blockx=16, blocky=16)# new fps: 20.2336

hopefully fixed now

Vapoursynth > Misc > Filter Order/Queue
Turn on QTGMC but mClean gets bolded instead.

Can't reproduce here. You might need to reset the 'Filter Order' in your defauts/configs.

(De-)Interlace/Telecine > VIVTC (Vapoursynth) > VFM
Errors only with telecined source, I believe:
> File "C:\Program Files\Hybrid\64bit\vsscripts\havsfunc.py", line 2271, in QTGMC_Interpolate
> eedi3 = partial(core.eedi3m.EEDI3CL, field=field, planes=planes, mdis=EdiMaxD, device=device, **eedi3_args)
> ^^^^^^^^^^^^^^^^^^^
> File "src\\cython\\vapoursynth.pyx", line 2691, in vapoursynth.Plugin.__getattr__
> AttributeError: There is no function named EEDI3CL
Changing Deinterlacer to something besides one of the QTGMC options works. Same issue with TIVTC

hopefully fixed now, same as the EEDI3CL problem above

In the deint tab, Auto-deint is set to None and [x] Overwrite... is greyed out (both checkbox and dropdown) and set to "telecine".
Fiddle with the deint settings until it shows [x] Overwrite... greyed out and set to "progressive" and it works again; very slow now.

hopefully fixed now, Overwrite should be enabled and set to 'progressive' is 'none' is selected.
If you do not want Hybrid to deinterlaced interlaced content, configure the encoder to output interlaced content.

And why is Auto-deint changing to match each new title load? I thought it usually just stayed at GTGMC. ]]
--> Restarting Hybrid now has Auto-deint back to QTGMC, even on progressive material, and [_] Overwrite... enabled but the dropdown disabled and set to "progressive".
This looks more normal and turning on the QTGMC filter in the VS tab now works fine.

Hybrid will not apply a deinterlacer if the scan type is not progressive and the encoder ist set to output interlaced content.

Other:

Filter Config Save saves/sets individual Deinterlace/Telecine settings... but not which Auto Deint Handling was selected.
Very weird contradiction. Don't know how to save this per job... and would prefer it saved/set with Filter Config.

Yes, this is as intended, depending on your source the auto deinterlace handling can change.

More of a feature add, I'm guessing: VIVTC/TIVTC/etc input and output frame counts are not correct for VFR sources (and could add a hint to manually set Change Speed [and/or fix that too]). Don't see a VFRtoCFR filter (or `ffmpeg -vsynch cfr` equivalent). Somehow this works without a CFR filter; not sure how. Oh, wait, I see AssumeFPS() in the script... but, no, that's not it since it contains the average VFR value. Anyway, I'm running into some confusion here that may or may not need work. Decimating telecined and interlaced VFR material _usually_ works well when I get the settings right... but it feels a bit like magic.

vfr content gets decoded as cfr when deinterlacing or frame reduction or interpolation gets applied
all too vage. no details -> nothing to look at

FDKAAC progress bar/text does not become visible until more than half done. This 20-second delay makes me poke around to make sure the job is running.

do vague

Hybrid writes APPDATA\hybrid\fontconfig\... folders in my working folder.

can't reproduce, need details

Audio cut short during detelecine (from a source flagged as telecine -- not an issue when decimating a source flagged as TFF). When setting audio to "custom" and encoding AC-3 to AAC FDK (no other codec tested), it cuts the audio short as if the length was changed by the number of frames the detelecine drops -- that is, it cuts off the last 1/5 of the audio track (presumably because it calcs the frames at 30fps still instead of 24fps). Tested "auto add (all)" and results were the same (but file size was 3kb [out of 50mb] larger somehow). Tested "auto add (all)" with multiple input files and IT WORKS!... very weird workaround, if you can call it that (I stumbled upon this by accident). I checked the temp *.aac file and it is also short, so not a muxing problem.

I can ivtc content without sync problems.
Share proper details. (read: https://www.selur.de/support)

The completion bar for the video during detelecine (including decimating while deinterlacing) hits 100% 4/5 of they way through and keeps going to 125%. Not sure if this is intentional or not. I can't see the audio bar due to the other issue that hides it for the first 20 seconds or so, at which point it is already done with this material.

Share proper details. (read: https://www.selur.de/support)

1920x1080 PAR 0.75 stretches Preview wrong way to 2560x1080 instead of 1440x1080 (and, yes, it was a weird to find such a source). I didn't try encoding it to see what happens. Instead, I forced a resize to 1440x1080 with 1:1 pixels before encode.

Share proper details and a source sample. (read: https://www.selur.de/support)

Bookmarks are lost whenever I reset the VS Preview. I would prefer it renamed the last tempPreviewVapoursynthFile*.vpy.bookmarks file to match the new tempPreviewVapoursynthFile*.vpy on refresh to keep the bookmarks.

patches for vsViewer (https://github.com/Selur/vsViewer) are welcome (I have never used the bookmarks in vsViewer, no clue what those are for)


Cu Selur

Ps.: send you a link to a dev version (via pm) for testing, link will be available ~24hrs.

[juanitogan]

Originally Posted by Selur

My guess is that you are running out of RAM that Avisynth can use.
Try lowering the thread count and if that does not help switch to 64bit.

The 'DePulse64' and 'KillPulse32' info should not cause problems.
This probably happens since you haven't: properly uninstalled Hybrid with settings, migrated your settings, reset your defaults when it was mentioned in the changelog

The first version of Hybrid I used was 2023.03.17.1 and I do uninstall between versions. I do not, however, uninstall the settings between versions. I read the changelog for all versions between my last and latest, but I guess I missed that somewhere -- that and resetting defaults. I might try all that when I get a chance, just to see what happens. The Avisynth interface is now a low priority since I now have AutoAdjust working in VS scripts. Also, I generally run Hybrid with 5-7gb of free memory, so I'm doubtful about RAM issues, but will double check that (and threads) regardless.

Originally Posted by Selur

I can ivtc content without sync problems.
Share proper details. (read: https://www.selur.de/support)

It's not a sync problem but a truncation problem. I'll do a proper debug on it and get back. It should be an obvious difference in the aac encode command line from one job versus another. I don't run into bonafide, non-interlaced, telecine DVD sources much, and the workaround to encode video separate from audio is obvious, but I'll try to get to this soon.

[Selur]

Also, I generally run Hybrid with 5-7gb of free memory,...

Avisynth 32bit is still limited to somewhat 3.5-4GB of RAM, so with a complex enough script and threads even with SD content you can hit that limit.

It's not a sync problem but a truncation problem.

Okay, I will need a debug output and probably a sample which allows to reproduce the problem.

Cu Selur

[Selur]

2024.05.11.1:
*changed*

• Vapoursynth: adjusted to DeOldify 0.3.5.2

*fixed*

• Vapoursynth: fixed RIFE + FillDrops, ReplaceSingle, FillDuplicates, ReplaceBlackFrames
• Vapoursynth: EEDI3/EEDI3CL handling
• Avisynth: Denoise filter not applied
• Avisynth: some typos in gui
• Avisynth: FrameNumber should not be in encode
• ProRes: progressive to interlaced encoding when no synth is us

-> downloads: http://www.selur.de/downloads

Cu Selur

[NoX1911]

I have Hybrid/20240511 and Vapoursynth installed (ArchLinux) but its not detected in 'Config/Tools/Tool Versions'. 'vspipe' is in $path. I cannot set it manually in 'Config/Tools' (there is no 'vspipe' slot). What has to be done here?

[Selur]

If you start Hybrid from a terminal, what is the output?
Is Vapoursynth support enabled?
Are vsPluginsPath,vsPluginsPath, vsScriptPath set inside a misc.ini ? (misc.ini can be a. next to Hybrid, b. inside ~/.hybrid c. inside .~/./config/hybrid)
Can vspipe be found via 'which vspipe' ?`

Cu Selur

[NoX1911]

Looks like the plugins folder doesn't get picked up so i've put it into ~/.hybrid/misc.ini.

[General]
vsPluginsPath=/usr/lib/vapoursynth
vsScriptPath=/home/user/ffmpeg/vapoursynth

After that, vapoursynth is enabled but doesn't run all too well (job crashes). Loading the script in vsedit shows:

2024-06-08 20:50:29.241
Failed to evaluate the script:
Python exception: module 'validate' has no attribute 'transferIsInvalid'

Traceback (most recent call last):
File "src/cython/vapoursynth.pyx", line 3365, in vapoursynth._vpy_evaluate
File "src/cython/vapoursynth.pyx", line 3366, in vapoursynth._vpy_evaluate
File "/tmp/encodingTempSynthSkript_2024-06-08@20_44_40_4310_0.vpy", line 20, in 
if validate.transferIsInvalid(clip):
^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: module 'validate' has no attribute 'transferIsInvalid'

Python version is 3.12.3.

The script is:

# Imports
import vapoursynth as vs
# getting Vapoursynth core
import sys
import os
core = vs.core
# Import scripts folder
scriptPath = '/home/user/ffmpeg/vapoursynth'
sys.path.insert(0, os.path.abspath(scriptPath))
# Import scripts
import validate
# Source: '/tmp/video.mp4'
# Current color space: YUV420P8, bit depth: 8, resolution: 480x640, frame rate: 29.848fps, scanorder: progressive, yuv luminance scale: limited, matrix: 470bg
# Loading /tmp/video.mp4 using LibavSMASHSource
clip = core.lsmas.LibavSMASHSource(source="/tmp/video.mp4")
frame = clip.get_frame(0)
# Setting detected color matrix (470bg).
clip = core.std.SetFrameProps(clip=clip, _Matrix=1)
# setting color transfer (709), if it is not set.
if validate.transferIsInvalid(clip):
  clip = core.std.SetFrameProps(clip=clip, _Transfer=1)
# setting color primaries info (to 709), if it is not set.
if validate.primariesIsInvalid(clip):
  clip = core.std.SetFrameProps(clip=clip, _Primaries=1)
# setting color range to TV (limited) range.
clip = core.std.SetFrameProps(clip=clip, _ColorRange=1)
# making sure frame rate is set to 29.848fps
clip = core.std.AssumeFPS(clip=clip, fpsnum=3731, fpsden=125)
# making sure the detected scan type is set (detected: progressive)
clip = core.std.SetFrameProps(clip=clip, _FieldBased=0) # progressive
# adjusting color space from YUV420P8 to RGB24 for vsLevels
clip = core.resize.Bicubic(clip=clip, format=vs.RGB24, matrix_in_s="470bg", range_s="limited")
# adjusting color using Levels on RGB24 (8 bit)
clip = core.std.Levels(clip=clip, min_in=16, max_in=235, min_out=16, max_out=100)
# adjusting output color from: RGB24 to YUV420P8 for x264Model
clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P8, matrix_s="470bg", range_s="limited")
# set output frame rate to 29.848fps (progressive)
clip = core.std.AssumeFPS(clip=clip, fpsnum=3731, fpsden=125)
# output
clip.set_output()

My plugins are:

$ ls -1 /usr/lib/vapoursynth
pkgconfig
bestsource.so
libbestsource.so
libf3kdb.so
libffms2.so
libvslsmashsource.so

I guess this is not hybrid ground anymore and i should better take this to Vapoursynth thread instead.

[Selur]

Not answering my questions, isn't really helping.
On Linux Hybrid does not use the pluginsPath in the script, since Python and Vapoursynth should be installed systemwide.

Python exception: module 'validate' has no attribute 'transferIsInvalid'

scriptPath = '/home/user/ffmpeg/vapoursynth'
sys.path.insert(0, os.path.abspath(scriptPath))

should add '/home/user/ffmpeg/vapoursynth' to the system path used to look for scripts.
Assuming that path includes https://github.com/Selur/Vapoursynth...er/validate.py Vapoursynth should find 'transferIsInvalid'.
Since Vapoursynth does not fail directly on 'import validate' the validate.py should be present.
=> One reason for the error could be that you got a different file named validate.py in that folder.

Cu Selur

[Selur]

2024.06.09.1:
*aded*

• Vapoursynth: torch, ProPainter

*changed*

• Vapoursynth: torch, adjusted to DeOldify 0.3.5.3
• Filtering: remove 'none' as auto deinterlacer handling option
• Video: NVEnc support --tflevel --lookahead-level

*fixed*

• Vapoursynth: another EEDI3/EEDI3CL handling
• Vapoursynth: Masking for RGBH filters
• QSVEnc: remove '--tier' during hevc encoding
• QSVEnc: set '--output-csp' (this is broken for most choices)
• Jobs: QSVEnc av1 encoding in webm
• Vapoursynth: torch, DeOldify fix FP16
• Vapoursynth: TFM/VFM, fps refresh
• Avisynth: custom with dual view
• Avisynth: interlaced output + cropView
• x265: 3.0 and 4.0 vbv values not set

-> downloads: http://www.selur.de/downloads

Cu Selur

[NoX1911]

Originally Posted by Selur

Not answering my questions, isn't really helping.

Sorry, here it is: hybrid-stdout.txt
That's from a pure AUR version that pulls the minimal version only (i didn't know that at that time). After realizing it needs more stuff (like the scripts) from the base/deps image it works properly now.
'vsPluginsPath' has nevertheless to be defined though. Hybrid doesn't look up /usr/lib/vapoursynth by itself and keeps VS disabled otherwise.

[Selur]

'vsPluginsPath' has nevertheless to be defined though.

Correct.

After realizing it needs more stuff (like the scripts) from the base/deps image it works properly now.

side note: the scripts Hybrid uses are present in hybrid\vsscripts
Sadly I have no clue how to make a full portable version with all the Python, scripts, filters included.
iirc, if you don't set vsScriptPath, Hybrid should find the scripts in (hybrid\vsscripts).
Hybrid also does try to find vsPluginsPath, but it does not look for it under ' /usr/lib/vapoursynth' but '/usr/local/lib/vapoursynth'. => all add ' /usr/lib/vapoursynth' to the default search path for the next release.

  './64bit/vsfilters' doesn't exist
  './vsfilters' doesn't exist
  '/home/user/.hybrid/vsfilters' doesn't exist
  '/home/user/opt/vapoursynth/vsplugins' doesn't exist
  '/home/user/opt/vapoursynth/lib/vapoursynth/' doesn't exist
  '/usr/local/lib/vapoursynth' doesn't exist
 '/home/user/.hybrid/vsscripts' doesn't exist,..
 './vsscripts' doesn't exist,..
 '/usr/share/vsscripts' doesn't exist,..
 '/home/user/opt/vapoursynth/vsscripts' doesn't exist,..

Cu Selur

[Selur]

2024.06.15.1:
*added*

• Vapoursynth: support CPreview from hello_hello

*changed*

• Vapoursynth: torch, adjust to DeOldify 4.0.0
• Vapoursynth: torch, adjust to vsRealESRGAN 5.0.0
• Vapoursynth: reworked FillDuplicateFrames, FillDrops, ReplaceSingle into one class for stability
• Vapoursynth: vs-mlrt, RIFE scale support optional requires 'onnx' in Vapoursynth/Lib/site-packages, see: https://github.com/AmusementClub/vs-mlrt/issues/94

*fixed*

• Vapoursynth: do not change color matrix on RGB content
• Vapoursynth: fixed some unnecessary, popups during startup
• Video: NVEnc support --tflevel --lookahead-level

-> downloads: http://www.selur.de/downloads

Cu Selur

Ps.: deinstall old version before installing this one.

[Selur]

2024.06.16.1:
*changed*

• Video: NVEnc, support '--vpn-nlmeans'

*fixed*

• Vapoursynth: initialization error on Windows

-> downloads: http://www.selur.de/downloads

Cu Selur

Ps.: This is a WIndows only release.

[Selur]

@Linux users:
Vapoursynth plugins build script is broken atm., got no time to look into it.
=> no new Hybrid Linux release until I find time to look into it (probably end of the year)

Cu Selur

Ps.: patches are welcomed https://github.com/Selur/hybrid-vapoursynth-addon/

[Selur]

*2024.09.29.1:
*changed*

• Vapoursynth: SoftLight require PC scale when mode != 8
• Vapoursynth: updated RIFE, RIFE(mlrt)
• Vapoursynth: use constants for Matrix, Primaries, Chroma, Luma
• Video: NVEnc/QSVEnc support --vpy-mt
• Jobs: add abort-option to size- and audio-warning onjo create&start
• Jobs: added scripts to report file
• Video: removed NVEnc input type checks
• Filtering: hide deinterlace control on progressive input, since it confuses users
• Avisynth: FFmpeg2Source use YUV422PX instead of YUY2
• Avisynth: increased suported memory max to 65536MB
• Avisynth: Hybrid now comes without Avisynth 32bit support, it is an addon now
• Video: do not check for UT Video, assume its present
• Vapoursynth: prefer BestSource as default for file input
• Vapoursynth: torch - added 'Requires'-controls to VSGAN and VSGAN Filter
• Input: Switch from Vapoursynth to Avisynth (when available) on .avs input
• Vapoursynth: TemporalDegrain2: degrainTR default and crash with degrainTR=0
• Video: x264, disable and deactivate 'auto' for color space on profile 'None'
• Video: x264, x264 - added 'adjust gop max to output fps'
• Ui: changed maximize handling
• Ui: option to set debug output path
• Vapoursynth: add masking for frame matching
• Vapoursynth: added FTF to TIVTC
• FFmpeg: don't set color signaling for pipes
• Avisynth: added 'Never use AviSource'

*fixed*

• Avisynth: enable avsToon for 32bit
• Video: x264 minimizeCL with 'superfast'
• Vapoursynth: FileQueue and Masking issues
• Vapoursynth: ReplaceSince
• Vapoursynth: torch - remove unnecessary resize call after VSGAN
• Extraction: use ffmpeg for ProRes from mkv extraction
• Video: SVT-AV1 typo 'enableResotration' vs 'enableRestoration'
• Vapoursynth: SMDegrain smdegrain.py script interlaced handling fixed
• Vapoursynth: mlrt - DPIR FP16 inverted
• Video: NVEnc - max bitrate initial value
• Video: NVEnc - EdgeLevel not added to command line
• Input: Zip format handling (ffmpeg and analysis)
• Synh: fix typo in CustomSynthScript QQDFPlus and abcxyz YUVALLX -> ALLYUVX
• Vapoursynth: QTGMC in CompareView with 'custom', use Preset='Fast' for original side.
• Vapoursynth: torch, VSGAN (Filter), width/height mixup
• Vapoursynth: torch, DeOldify 'render_vivid' and 'dark' passed as string instead of bool
• Vapoursynth: SmoothGrad fix min value 1
• Vapoursynth: Frame->Resize UI scaling
• Vapoursynth: torch, ProPainter 'Inference'-CheckBox not working
• Vapoursynth: cfhd handling
• Vapoursynth: SSIMD: fix output color
• Vapoursynth: letterboxing YUV422 mod2
• Vapoursynth: vsRemoveDirt defaults
• Video: FFmpegs zscale-call for PNG output
• Video: x264 slow preset using b-adapt 2 instead of 1
• FFmpeg: 2020 signaling
• Extract: vp9 video
• Model: loadfromJson respect skipList
• Avisynth: TIVTC+QTGMC
• Avisynth: fix histogram color restriction check, no YUY2 support
• Avisynth: QTGMC+Custom+UltraFast missing YadifMod2 dependency
• Avisynth: DeblockPP7 and aSharp were unnecessarily disabled for 64bit Avisynth
• Video: ProRes encoding when mencoder is used as decoder
• Vapoursynth: DecimateMode set output fps for mode 2, which is (still) broken
• Vapoursynth: RainbowSmooth require mod16
• Vapoursynth: don't try to set ColorMatrix on RGB content
• Vapoursynth: DGSharpen requires mod16
• Vapoursynth: rectangular mask for high bit depth
• Decoder: changed Avisynth input handling
• Muxing: mkv helper for ffmpeg color matrix and vfr
• Jobs: fix cleanup job handling on raw audio only output
• Jobs: fix warning handling

*added*

• Vapoursynth: GLSL Gamma, RGBAdjust filter ports(Levels broken&disabled)
• Vapoursynth: proToon
• Vapoursynth: vs_temporalfix
• Vapoursynth: torch/vmlrt option to set (trt) 'engine path'
• Avisynth: added CustomSynthScripts for 32bit and 64bit proToon-v0.75.1
• Generate output name suffix added '%DATE%','%TIME%','%TIMEMS%' replacements
• Vapoursynth: torch - add vs_colorfix for VSGAN and VSGANFilter

-> downloads: http://www.selur.de/downloads

Cu Selur

[Leo 69]

Just installed the last version of Hybrid, loaded a movie (regular H.264 video in .mkv container) for re-encoding, then navigated to "Filtering" section and without enabling any Vapoursynth filters, pressed the "Preview" button ========> Nothing happened, and no preview window appeared. When a couple of filters are enabled, the preview window doesn't appear either.

Is this a designed behavior? How to fix it and make the preview window appear so that I could actually see the effect of applied filters? I've attached the debug log.

https://drive.filen.io/d/228e5b58-7a...Aeq7LGIrMkRXjS

[Selur]

Probably a problem with BestSource, disabling "Filtering->Vapoursynth->Misc->Source->Prefer BestSource" or replacing the Hybrid\64bit\vsfilters\SourceFilter\BestSource\BestSource.dll with R6 from https://github.com/vapoursynth/bestsource/releases should help.
Would need a sample of your source to say for sure where the problem is.

Cu Selur

[Leo 69]

Selur,

Thank you very much. Eventually I found that ffms2 is working better for my use case and I can see the preview window now.

Could I ask you some questions?

How to make this work in Hybrid? => https://github.com/JingyunLiang/RVRT

Is converting from .pth to .onnx with Chainner sufficient to make this work? How to import these models into Hybrid? Are there models similar but better that these?

And also does Hybrid support importing .webp or .avif image sequences as input?

Also, I have say that Hybrid is the most complex and comprehensive encoding tool I've used. The amount of mental effort that was put into it by just one person is incredible.