https://forum.selur.net/ Wed, 03 Jun 2026 05:38:58 +0000 MyBB https://forum.selur.net/thread-4393.html Fri, 29 May 2026 11:12:27 +0200 bigOrange123]]> https://forum.selur.net/thread-4393.html I'm restoring old films and TV series and keep running into a recurring image quality issue: strong edges often show a purple band or halo. The artifact appears consistently around candlelight in night scenes, as well as along high-contrast edges in daylight. Here are a few examples
     

I'm not entirely sure of the correct term – it may be related to “purple flare” or chromatic aberration. It’s very distracting and seriously degrades the viewing experience.
 
The same clips frequently suffer from a related dark edge problem. For example, candle flames often show a dark line on one side and a purple fringe on the other – extremely annoying.


I'm looking for an effective method/workflow that can remove these artifacts. I will provide a short test video; you only need to process the segments shown in the screenshots (no need to run the entire clip).

test video clip
https://drive.google.com/file/d/1kmni1sN...sp=sharing

Key requirements:


Minimize luma (Y) channel processing as much as possible, and avoid noticeable detail loss. Ideally, the Y channel should be left mostly untouched, but limited processing is acceptable if it doesn’t degrade fine details significantly. This step serves as a pre-processing module before an AI model handles detail restoration, so preserving image information is critical.

Please note: the test source has been badly upscaled (not by me – I received it this way). You are free to downscale it if necessary.

If your solution meets my requirements, I'll pay $50 USD. Please provide a clear workflow (script, filter chain, or detailed steps) that I can apply. ]]> I'm restoring old films and TV series and keep running into a recurring image quality issue: strong edges often show a purple band or halo. The artifact appears consistently around candlelight in night scenes, as well as along high-contrast edges in daylight. Here are a few examples
     

I'm not entirely sure of the correct term – it may be related to “purple flare” or chromatic aberration. It’s very distracting and seriously degrades the viewing experience.
 
The same clips frequently suffer from a related dark edge problem. For example, candle flames often show a dark line on one side and a purple fringe on the other – extremely annoying.


I'm looking for an effective method/workflow that can remove these artifacts. I will provide a short test video; you only need to process the segments shown in the screenshots (no need to run the entire clip).

test video clip
https://drive.google.com/file/d/1kmni1sN...sp=sharing

Key requirements:


Minimize luma (Y) channel processing as much as possible, and avoid noticeable detail loss. Ideally, the Y channel should be left mostly untouched, but limited processing is acceptable if it doesn’t degrade fine details significantly. This step serves as a pre-processing module before an AI model handles detail restoration, so preserving image information is critical.

Please note: the test source has been badly upscaled (not by me – I received it this way). You are free to downscale it if necessary.

If your solution meets my requirements, I'll pay $50 USD. Please provide a clear workflow (script, filter chain, or detailed steps) that I can apply. ]]> https://forum.selur.net/thread-4392.html Fri, 29 May 2026 06:06:59 +0200 Oldermediaformats]]> https://forum.selur.net/thread-4392.html
When I’ve clicked on the Vapoursynth preview, a new window opens with that in the display, but after a few seconds, it crashes out (although gui of Hybrid remains open).

When I click on unins000 icon to perform the uninstall, it asks me if I want to remove my settings too. I have clicked No. I reinstall Hybrid and restart my laptop, but when I open Hybrid again with a new job, it continues to crash.

I have contemplated a thorough uninstall using Revo and conducting a thorough scan wherein I could clear all registry items. If I go that route, I’ve saved a backup of the scripts subfolder containing profiles of Avisynth and Vapoursynth with my custom filters preserved.

I’m just perplexed how I can get it running smoothly again.]]>
When I’ve clicked on the Vapoursynth preview, a new window opens with that in the display, but after a few seconds, it crashes out (although gui of Hybrid remains open).

When I click on unins000 icon to perform the uninstall, it asks me if I want to remove my settings too. I have clicked No. I reinstall Hybrid and restart my laptop, but when I open Hybrid again with a new job, it continues to crash.

I have contemplated a thorough uninstall using Revo and conducting a thorough scan wherein I could clear all registry items. If I go that route, I’ve saved a backup of the scripts subfolder containing profiles of Avisynth and Vapoursynth with my custom filters preserved.

I’m just perplexed how I can get it running smoothly again.]]> https://forum.selur.net/thread-4391.html Mon, 25 May 2026 11:36:32 +0200 Bartoloni]]> https://forum.selur.net/thread-4391.html


and a video
http://www.wcn.it/h.mp4]]>


and a video
http://www.wcn.it/h.mp4]]> https://forum.selur.net/thread-4390.html Sun, 24 May 2026 20:21:05 +0200 Oldermediaformats]]> https://forum.selur.net/thread-4390.html
I also need to get the source's 29.97 fps converted to 23.976. Should I check any of the filters in the Speed Change tab and then select my desired frame rate?

I will eventually resize file's resolution from 720x480 to 1280x720.]]>
I also need to get the source's 29.97 fps converted to 23.976. Should I check any of the filters in the Speed Change tab and then select my desired frame rate?

I will eventually resize file's resolution from 720x480 to 1280x720.]]> https://forum.selur.net/thread-4389.html Sat, 23 May 2026 08:27:00 +0200 andrewschen]]> https://forum.selur.net/thread-4389.html VapoursynthR73_torch_2026.02.26 / VapoursynthR76_torch_2026.05.16 both tried
vs-mlrt_2026.03.18 / vs-mlrt_2026.03.26 both tried

encode FHD 1920x1080 video, svt-av1 crf20 preset 6 temporal filter 0 with 2~4 fps

original speed (no vs filter) is up to 80+ fps
Intel 275hx / 5090 laptop / 596.36 studio driver
3A game runs fine

Will test Hybrid_dev_2026.05.22-18618 later

Is there any setting I should do with geforce driver or nvidia app?

  HybridDebugOutput.zip (Size: 25,7 KB / Downloads: 16) ]]> VapoursynthR73_torch_2026.02.26 / VapoursynthR76_torch_2026.05.16 both tried
vs-mlrt_2026.03.18 / vs-mlrt_2026.03.26 both tried

encode FHD 1920x1080 video, svt-av1 crf20 preset 6 temporal filter 0 with 2~4 fps

original speed (no vs filter) is up to 80+ fps
Intel 275hx / 5090 laptop / 596.36 studio driver
3A game runs fine

Will test Hybrid_dev_2026.05.22-18618 later

Is there any setting I should do with geforce driver or nvidia app?

  HybridDebugOutput.zip (Size: 25,7 KB / Downloads: 16) ]]> https://forum.selur.net/thread-4388.html Fri, 22 May 2026 17:19:34 +0200 georgepriftakis]]> https://forum.selur.net/thread-4388.html

https://mega.nz/file/0MYHTaaJ#KRsD7zwtik...v5Eyh_rLfM]]>

https://mega.nz/file/0MYHTaaJ#KRsD7zwtik...v5Eyh_rLfM]]> https://forum.selur.net/thread-4386.html Thu, 21 May 2026 09:35:45 +0200 Bartoloni]]> https://forum.selur.net/thread-4386.html but on my case any footage is always 1:1 (from OBS) and i need to change to 16:15 for every JOB ( the cropping is saved correctly and lookly i don't need to edit these 8 parameters) and force with the PAR FLAG



can be nice if when saving as default parameter also this flag and these 2 parameters can be saved . (or add an extra flag to the PIXEL ASPECT RATIO form to OVERRIDE any automatically filled parameter?)

Have a nice day Selur.]]> but on my case any footage is always 1:1 (from OBS) and i need to change to 16:15 for every JOB ( the cropping is saved correctly and lookly i don't need to edit these 8 parameters) and force with the PAR FLAG



can be nice if when saving as default parameter also this flag and these 2 parameters can be saved . (or add an extra flag to the PIXEL ASPECT RATIO form to OVERRIDE any automatically filled parameter?)

Have a nice day Selur.]]> https://forum.selur.net/thread-4385.html Sat, 16 May 2026 22:15:51 +0200 NASS]]> https://forum.selur.net/thread-4385.html What you’ve done with Hybrid is truly magnificent.
I wanted to report an issue regarding ColorAdjust with the ReColor option. I’m using the official Hybrid version (Hybrid_2026.03.21.1_SETUP), not the development build and VapoursynthR73_torch_2026.02.26.
The issue that still persists is significant distortions on moving objects when ReColor is enabled. I tried reducing the problem using:

qtgmc.QTGMC(clip, Preset="Very Slow", InputType=3, TR2=2, TFF=False, SourceMatch=0, Lossless=0)

and it reduced the issue a little, but the distortions are still visible.
It’s a shame because this is a really great feature, but this small problem affects the final result. We definitely need to find a solution!

https://drive.google.com/drive/folders/1...KwSv1D5ifD

PS: P.S. In the development version ColorAdjust with the ReColor, there is also a temporal consistency issue across the whole image, similar to a flickering effect from frame to frame + Distortion .

Best regards]]> What you’ve done with Hybrid is truly magnificent.
I wanted to report an issue regarding ColorAdjust with the ReColor option. I’m using the official Hybrid version (Hybrid_2026.03.21.1_SETUP), not the development build and VapoursynthR73_torch_2026.02.26.
The issue that still persists is significant distortions on moving objects when ReColor is enabled. I tried reducing the problem using:

qtgmc.QTGMC(clip, Preset="Very Slow", InputType=3, TR2=2, TFF=False, SourceMatch=0, Lossless=0)

and it reduced the issue a little, but the distortions are still visible.
It’s a shame because this is a really great feature, but this small problem affects the final result. We definitely need to find a solution!

https://drive.google.com/drive/folders/1...KwSv1D5ifD

PS: P.S. In the development version ColorAdjust with the ReColor, there is also a temporal consistency issue across the whole image, similar to a flickering effect from frame to frame + Distortion .

Best regards]]> https://forum.selur.net/thread-4384.html Fri, 15 May 2026 19:27:39 +0200 RetroHelix]]> https://forum.selur.net/thread-4384.html


Now I was wondering if I can pimp the output a bit and tried to remove rainbow effects (see the lapel of the right guy and the necktie (green at the end of the screen) of the left guy) but was not able to get a good output:

The effect is even stronger than in the original :\

I tried some suggestions from here (link) but Hqdn3d and/or Degraining was not helping. Honestly I feel a bit lost. I'm just too dumb to alter the parameters of the filters in a meaningful way.

In the original Video you can see heavy rainbowing in the beginning and at 1:30.

Would be nice if someone is able to help me with this. Can this even be removed?


Thank you]]>


Now I was wondering if I can pimp the output a bit and tried to remove rainbow effects (see the lapel of the right guy and the necktie (green at the end of the screen) of the left guy) but was not able to get a good output:

The effect is even stronger than in the original :\

I tried some suggestions from here (link) but Hqdn3d and/or Degraining was not helping. Honestly I feel a bit lost. I'm just too dumb to alter the parameters of the filters in a meaningful way.

In the original Video you can see heavy rainbowing in the beginning and at 1:30.

Would be nice if someone is able to help me with this. Can this even be removed?


Thank you]]> https://forum.selur.net/thread-4383.html Wed, 13 May 2026 15:40:41 +0200 Miranda]]> https://forum.selur.net/thread-4383.html
Any suggestions?]]>
Any suggestions?]]> https://forum.selur.net/thread-4382.html Tue, 12 May 2026 13:34:52 +0200 Fjorse]]> https://forum.selur.net/thread-4382.html  
ich nutze Hybrid seit ca. einem Jahr um alte VHS-Aufnahmen in HD bei 50 FPS zu konvertieren. Das hat auch immer gut funktioniert. Mit der aktuellen Version 2026.03.21.1 bekomme ich gleich am Anfang des "Jobs" den Fehler "crashed - encoding video to prores using ffmpeg status 0".
Daten Quelldatei:
- Capture mit Blackmagic Intensity Shuttle (Quicktime losless 8 Bit)
- YUV 4:2:2 
- 25FPS/ Halbbilder

Die Datei wird sauber eingelesen und ich kann meine Optionen (-> Vapoursynth) setzen:
- Auflösung in HD
- 50 FPS (Bob)
- Color-Matrix
- DeHalo

Wenn ich den Job starte, bricht der Prozess bei ca. 14% ab. Wenn ich das ganze mit AVISynth versuche, scheint es zu funktionieren - habe aber vorzeitig abgebrochen, da mir zu viel Zeit verloren geht und ich ja Vapoursynth verwenden möchte.
Was hat sich zwischen den Versionen geändert, weshalb es nicht mehr funktioniert? Hinweis: ich habe die neue Version "drüberinstalliert" -> unsauber?

Gruß, Frank]]>  
ich nutze Hybrid seit ca. einem Jahr um alte VHS-Aufnahmen in HD bei 50 FPS zu konvertieren. Das hat auch immer gut funktioniert. Mit der aktuellen Version 2026.03.21.1 bekomme ich gleich am Anfang des "Jobs" den Fehler "crashed - encoding video to prores using ffmpeg status 0".
Daten Quelldatei:
- Capture mit Blackmagic Intensity Shuttle (Quicktime losless 8 Bit)
- YUV 4:2:2 
- 25FPS/ Halbbilder

Die Datei wird sauber eingelesen und ich kann meine Optionen (-> Vapoursynth) setzen:
- Auflösung in HD
- 50 FPS (Bob)
- Color-Matrix
- DeHalo

Wenn ich den Job starte, bricht der Prozess bei ca. 14% ab. Wenn ich das ganze mit AVISynth versuche, scheint es zu funktionieren - habe aber vorzeitig abgebrochen, da mir zu viel Zeit verloren geht und ich ja Vapoursynth verwenden möchte.
Was hat sich zwischen den Versionen geändert, weshalb es nicht mehr funktioniert? Hinweis: ich habe die neue Version "drüberinstalliert" -> unsauber?

Gruß, Frank]]> https://forum.selur.net/thread-4381.html Sat, 02 May 2026 19:03:43 +0200 r3dfx]]> https://forum.selur.net/thread-4381.html

[color=#800000][font=Segoe UI]Failed to get VSScript API![/font][/color] [color=#800000][font=Segoe UI]Failed to create VSScript handle![/font][/color]

Probably the title is somewhat misleading, but that was the initial issue

Debug output

I've made a suboptimal decision of not keeping original .exe to use as a fallback solution...

UPD: looks like Hybrid cooked my windows install somehow
This is what the end of debug file looks now when I try to open any preview. And this is on previously stable 2025.11.09.1

2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Preview::indexingIsFinished 'VAPOURSYNTHPREVIEW' ignored,.. 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Got unhandled IPC message: Failed to initialize VapourSynth environment! 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Got unhandled IPC message: Failed to initialize VapourSynth environment! 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Got unhandled IPC message: Failed to initialize VapourSynth environment! 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Got unhandled IPC message: Failed to initialize VapourSynth environment! 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Failed to evaluate the script.

]]>

[color=#800000][font=Segoe UI]Failed to get VSScript API![/font][/color] [color=#800000][font=Segoe UI]Failed to create VSScript handle![/font][/color]

Probably the title is somewhat misleading, but that was the initial issue

Debug output

I've made a suboptimal decision of not keeping original .exe to use as a fallback solution...

UPD: looks like Hybrid cooked my windows install somehow
This is what the end of debug file looks now when I try to open any preview. And this is on previously stable 2025.11.09.1

2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Preview::indexingIsFinished 'VAPOURSYNTHPREVIEW' ignored,.. 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Got unhandled IPC message: Failed to initialize VapourSynth environment! 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Got unhandled IPC message: Failed to initialize VapourSynth environment! 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Got unhandled IPC message: Failed to initialize VapourSynth environment! 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Got unhandled IPC message: Failed to initialize VapourSynth environment! 2026.05.02 - 21:21:18_Windows 10 Version 21H2 (64bit)_2025.11.09.1 - level 9: Failed to evaluate the script.

]]> https://forum.selur.net/thread-4380.html Fri, 01 May 2026 21:23:59 +0200 r3dfx]]> https://forum.selur.net/thread-4380.html

---

Error message:

File "src\\cython\\vapoursynth.pyx", line 3123, in vapoursynth.Function.__call__ vapoursynth.Error: Analyse: failed to retrieve first frame from super clip. Error message: Resize error 1027: image dimensions must be divisible by subsampling factor

---

What was done: I've inserted # requires mod 2 before QTGMC. Same behavior, but padding is visible in the generated .vpy script 

---

A snippet from .vpy generated script:

# Current color space: YUV422P10, bit depth: 10, resolution: 720x486, frame rate: 29.97fps, scanorder: top field first, yuv luminance scale: limited, matrix: 170m, transfer: bt.709, primaries: bt.601 ntsc, format: prores # Loading I:\kseniya_tape1_ntsc.mov using BestSource) clip = core.bs.VideoSource(source="I:/kseniya_tape1_ntsc.mov", cachepath="I:/TEMP/kseniya_tape1_ntsc_bestSource", track=0, hwdevice="cuda") frame = clip.get_frame(0) # setting color matrix to 170m. clip = core.std.SetFrameProps(clip, _Matrix=vs.MATRIX_ST170_M) # setting color transfer (vs.TRANSFER_BT709), if it is not set. if validate.transferIsInvalid(clip): clip = core.std.SetFrameProps(clip=clip, _Transfer=vs.TRANSFER_BT709) # setting color primaries info (to vs.PRIMARIES_BT470_BG), if it is not set. if validate.primariesIsInvalid(clip): clip = core.std.SetFrameProps(clip=clip, _Primaries=vs.PRIMARIES_BT470_BG) # setting color range to TV (limited) range. clip = core.std.SetFrameProps(clip=clip, _ColorRange=vs.RANGE_LIMITED) # making sure frame rate is set to 29.97fps clip = core.std.AssumeFPS(clip=clip, fpsnum=30000, fpsden=1001) # Converting to target color sampling clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P8, dither_type="error_diffusion") # making sure the detected scan type is set (detected: top field first) clip = core.std.SetFrameProps(clip=clip, _FieldBased=vs.FIELD_TOP) # tff original = clip # adjusting color format to YUV422P8 for custom section clip = core.resize.Bicubic(clip=clip, format=vs.YUV422P8) [clip] = beforeDeinterlace(clip) # clip current meta; color space: YUV422P8, bit depth: 8, resolution: 720x486, fps: 29.97, color matrix: 170m, yuv luminance scale: limited, scanorder: top field first, full height: true clip = core.std.AddBorders(clip=clip, left=0, right=0, top=0, bottom=2) # add borders to archive mod 4 (vsQTGMC) - 720x488 # Deinterlacing using QTGMC clip = qtgmc.QTGMC(Input=clip, Preset="Medium", InputType=0, TFF=True, TR2=0, SourceMatch=1, Lossless=0, opencl=True) # new fps: 59.94 clip = core.std.Crop(clip=clip, left=0, right=0, top=0, bottom=2) # removing added borders from mod requirement (vsQTGMC) - 720x486 # Making sure content is preceived as frame based clip = core.std.SetFrameProps(clip=clip, _FieldBased=vs.FIELD_PROGRESSIVE) # progressive original = core.std.AddBorders(clip=original, left=0, right=0, top=0, bottom=2) # add borders to archive mod 4 (vsQTGMC) - 720x488 # Deinterlacing using QTGMC original = qtgmc.QTGMC(original, Preset="Fast", TFF=True, opencl=True)# new fps: 59.94 original = core.std.Crop(clip=original, left=0, right=0, top=0, bottom=2) # removing added borders from mod requirement (vsQTGMC) - 720x486 # Making sure content is preceived as frame based original = core.std.SetFrameProps(clip=original, _FieldBased=vs.FIELD_PROGRESSIVE) # progressive # adjusting output color from: YUV422P8 to YUV420P8 for NVEncModel clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P8) # add histogram clip = core.hist.Classic(clip) original = core.hist.Classic(original) original = core.text.Text(clip=original,text="Original",scale=1,alignment=7) clip = core.text.Text(clip=clip,text="Filtered",scale=1,alignment=7) stacked = core.std.StackHorizontal([original,clip]) # adjusted resizing to achieve PAR 1:1 for preview # resizing for preview to square pixel to 1310x742 stacked = core.resize.Bicubic(clip=stacked, width=1310, height=742) # set output frame rate to 59.94fps (progressive) stacked = core.std.AssumeFPS(clip=stacked, fpsnum=60000, fpsden=1001)

---

Hybrid 2025.11.09.1 was used, I've never ran into such issue before]]>

---

Error message:

File "src\\cython\\vapoursynth.pyx", line 3123, in vapoursynth.Function.__call__ vapoursynth.Error: Analyse: failed to retrieve first frame from super clip. Error message: Resize error 1027: image dimensions must be divisible by subsampling factor

---

What was done: I've inserted # requires mod 2 before QTGMC. Same behavior, but padding is visible in the generated .vpy script 

---

A snippet from .vpy generated script:

# Current color space: YUV422P10, bit depth: 10, resolution: 720x486, frame rate: 29.97fps, scanorder: top field first, yuv luminance scale: limited, matrix: 170m, transfer: bt.709, primaries: bt.601 ntsc, format: prores # Loading I:\kseniya_tape1_ntsc.mov using BestSource) clip = core.bs.VideoSource(source="I:/kseniya_tape1_ntsc.mov", cachepath="I:/TEMP/kseniya_tape1_ntsc_bestSource", track=0, hwdevice="cuda") frame = clip.get_frame(0) # setting color matrix to 170m. clip = core.std.SetFrameProps(clip, _Matrix=vs.MATRIX_ST170_M) # setting color transfer (vs.TRANSFER_BT709), if it is not set. if validate.transferIsInvalid(clip): clip = core.std.SetFrameProps(clip=clip, _Transfer=vs.TRANSFER_BT709) # setting color primaries info (to vs.PRIMARIES_BT470_BG), if it is not set. if validate.primariesIsInvalid(clip): clip = core.std.SetFrameProps(clip=clip, _Primaries=vs.PRIMARIES_BT470_BG) # setting color range to TV (limited) range. clip = core.std.SetFrameProps(clip=clip, _ColorRange=vs.RANGE_LIMITED) # making sure frame rate is set to 29.97fps clip = core.std.AssumeFPS(clip=clip, fpsnum=30000, fpsden=1001) # Converting to target color sampling clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P8, dither_type="error_diffusion") # making sure the detected scan type is set (detected: top field first) clip = core.std.SetFrameProps(clip=clip, _FieldBased=vs.FIELD_TOP) # tff original = clip # adjusting color format to YUV422P8 for custom section clip = core.resize.Bicubic(clip=clip, format=vs.YUV422P8) [clip] = beforeDeinterlace(clip) # clip current meta; color space: YUV422P8, bit depth: 8, resolution: 720x486, fps: 29.97, color matrix: 170m, yuv luminance scale: limited, scanorder: top field first, full height: true clip = core.std.AddBorders(clip=clip, left=0, right=0, top=0, bottom=2) # add borders to archive mod 4 (vsQTGMC) - 720x488 # Deinterlacing using QTGMC clip = qtgmc.QTGMC(Input=clip, Preset="Medium", InputType=0, TFF=True, TR2=0, SourceMatch=1, Lossless=0, opencl=True) # new fps: 59.94 clip = core.std.Crop(clip=clip, left=0, right=0, top=0, bottom=2) # removing added borders from mod requirement (vsQTGMC) - 720x486 # Making sure content is preceived as frame based clip = core.std.SetFrameProps(clip=clip, _FieldBased=vs.FIELD_PROGRESSIVE) # progressive original = core.std.AddBorders(clip=original, left=0, right=0, top=0, bottom=2) # add borders to archive mod 4 (vsQTGMC) - 720x488 # Deinterlacing using QTGMC original = qtgmc.QTGMC(original, Preset="Fast", TFF=True, opencl=True)# new fps: 59.94 original = core.std.Crop(clip=original, left=0, right=0, top=0, bottom=2) # removing added borders from mod requirement (vsQTGMC) - 720x486 # Making sure content is preceived as frame based original = core.std.SetFrameProps(clip=original, _FieldBased=vs.FIELD_PROGRESSIVE) # progressive # adjusting output color from: YUV422P8 to YUV420P8 for NVEncModel clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P8) # add histogram clip = core.hist.Classic(clip) original = core.hist.Classic(original) original = core.text.Text(clip=original,text="Original",scale=1,alignment=7) clip = core.text.Text(clip=clip,text="Filtered",scale=1,alignment=7) stacked = core.std.StackHorizontal([original,clip]) # adjusted resizing to achieve PAR 1:1 for preview # resizing for preview to square pixel to 1310x742 stacked = core.resize.Bicubic(clip=stacked, width=1310, height=742) # set output frame rate to 59.94fps (progressive) stacked = core.std.AssumeFPS(clip=stacked, fpsnum=60000, fpsden=1001)

---

Hybrid 2025.11.09.1 was used, I've never ran into such issue before]]> https://forum.selur.net/thread-4379.html Sat, 25 Apr 2026 20:12:51 +0200 NASS]]> https://forum.selur.net/thread-4379.html I would like to open a discussion about an important challenge for exemplar-based colorization methods:
Keyframe Color Consistency Challenge
We need to find a reliable way to achieve temporal consistency between colorized keyframes so they can be used for the full colorization of black and white videos.
At the moment, one of the main limitations is that even when keyframes are individually colorized well, colors often vary from one keyframe to another. This creates flickering, unstable tones, skin tone changes, clothing color drift, and inconsistencies in objects or environments when processing the full sequence.
Why This Matters
For real-world video restoration, getting beautiful frame-by-frame results is not enough. We need:

• Stable skin tones from scene to scene
  
• Consistent clothing and object colors
  
• Lighting adaptation without color drift
  
• Smooth transitions between distant keyframes
  
• Reliable propagation across long sequences
  

Some Approaches I Tested
I experimented with several workflows using Automatic1111 + ControlNet + IP-Adapter.
Example Setup in Automatic1111
Main txt2img Prompt
Positive Prompt:

masterpiece, best quality, colorized photograph, vibrant, highly detailed

Negative Prompt:

black and white, grayscale, monochrome, bad quality, deformed

ControlNet Unit 0 (Preserve B&W Structure)

• Enable: Yes
  
• Pixel Perfect: Yes
  
• Input Image: Original black and white frame
  
• Control Type: Recolor
  
• Preprocessor:
  
  recolor_luminance
  or
  
  recolor_intensity
  
  
• Model:
  
  ioclab_sd15_recolor
  
  
• Weight:
  
  1.0
  (or
  
  0.8
  if too strong)
  

ControlNet Unit 1 (Inject Keyframe Colors)

• Enable: Yes
  
• Pixel Perfect: Yes
  
• Input Image: Previously colorized reference keyframe
  
• Control Type: IP-Adapter
  
• Preprocessor:
  
  ip-adapter_clip_sd15
  
  
• Model:
  
  ip-adapter_sd15
  
  
• Weight:
  
  0.8 – 1.0
  
  

This helps transfer colors from one keyframe to another while respecting the luminance of the target black and white frame.
But Problems Remain
Even with this method:

• Colors still drift over time
  
• Different scenes generate different palettes
  
• The same person may change skin tone
  
• Sky, grass, and buildings vary between shots
  
• Propagation becomes unstable in long videos
  

2 - Qwen-Image-Edit
From my tests, Qwen-Image-Edit currently delivers some of the best single-keyframe colorization results in terms of realism, detail, and natural palette generation.
However, there is still no practical way to maintain the same temporal consistency across all other keyframes in the video.
So even if one keyframe looks excellent, the next keyframe may still shift in:

• Skin tone
  
• Clothing colors
  
• Environment palette
  
• Object consistency
  
• Scene mood
  

Main Question
Could it be possible to use Qwen-Image-Edit for colorize all keyframes?
For example:

• Use one master keyframe style and propagate it
  
• Multi-keyframe palette memory
  
• Identity/object color locking
  
• Reference-aware batch processing
  
• Latent temporal consistency between generated keyframes
  

What Do You Think?
Is there a way to use Qwen-Image-Edit to colorize all keyframes and create one unified temporal color source for all reference images (keyframes)?
This could be a major breakthrough for AI video colorization.
Your turn — ideas, experiments, thoughts?]]> I would like to open a discussion about an important challenge for exemplar-based colorization methods:
Keyframe Color Consistency Challenge
We need to find a reliable way to achieve temporal consistency between colorized keyframes so they can be used for the full colorization of black and white videos.
At the moment, one of the main limitations is that even when keyframes are individually colorized well, colors often vary from one keyframe to another. This creates flickering, unstable tones, skin tone changes, clothing color drift, and inconsistencies in objects or environments when processing the full sequence.
Why This Matters
For real-world video restoration, getting beautiful frame-by-frame results is not enough. We need:

• Stable skin tones from scene to scene
  
• Consistent clothing and object colors
  
• Lighting adaptation without color drift
  
• Smooth transitions between distant keyframes
  
• Reliable propagation across long sequences
  

Some Approaches I Tested
I experimented with several workflows using Automatic1111 + ControlNet + IP-Adapter.
Example Setup in Automatic1111
Main txt2img Prompt
Positive Prompt:

masterpiece, best quality, colorized photograph, vibrant, highly detailed

Negative Prompt:

black and white, grayscale, monochrome, bad quality, deformed

ControlNet Unit 0 (Preserve B&W Structure)

• Enable: Yes
  
• Pixel Perfect: Yes
  
• Input Image: Original black and white frame
  
• Control Type: Recolor
  
• Preprocessor:
  
  recolor_luminance
  or
  
  recolor_intensity
  
  
• Model:
  
  ioclab_sd15_recolor
  
  
• Weight:
  
  1.0
  (or
  
  0.8
  if too strong)
  

ControlNet Unit 1 (Inject Keyframe Colors)

• Enable: Yes
  
• Pixel Perfect: Yes
  
• Input Image: Previously colorized reference keyframe
  
• Control Type: IP-Adapter
  
• Preprocessor:
  
  ip-adapter_clip_sd15
  
  
• Model:
  
  ip-adapter_sd15
  
  
• Weight:
  
  0.8 – 1.0
  
  

This helps transfer colors from one keyframe to another while respecting the luminance of the target black and white frame.
But Problems Remain
Even with this method:

• Colors still drift over time
  
• Different scenes generate different palettes
  
• The same person may change skin tone
  
• Sky, grass, and buildings vary between shots
  
• Propagation becomes unstable in long videos
  

2 - Qwen-Image-Edit
From my tests, Qwen-Image-Edit currently delivers some of the best single-keyframe colorization results in terms of realism, detail, and natural palette generation.
However, there is still no practical way to maintain the same temporal consistency across all other keyframes in the video.
So even if one keyframe looks excellent, the next keyframe may still shift in:

• Skin tone
  
• Clothing colors
  
• Environment palette
  
• Object consistency
  
• Scene mood
  

Main Question
Could it be possible to use Qwen-Image-Edit for colorize all keyframes?
For example:

• Use one master keyframe style and propagate it
  
• Multi-keyframe palette memory
  
• Identity/object color locking
  
• Reference-aware batch processing
  
• Latent temporal consistency between generated keyframes
  

What Do You Think?
Is there a way to use Qwen-Image-Edit to colorize all keyframes and create one unified temporal color source for all reference images (keyframes)?
This could be a major breakthrough for AI video colorization.
Your turn — ideas, experiments, thoughts?]]> https://forum.selur.net/thread-4378.html Fri, 24 Apr 2026 22:28:10 +0200 davsed]]> https://forum.selur.net/thread-4378.html I'm trying to get DPIR (via vs-mlrt) working in Hybrid so I can denoise on my Intel Lunar Lake iGPU through the OpenVINO backend. The filter shows up in Hybrid's dropdown but I can't actually configure or run it. Hoping you can point me in the right direction.

System:

• OS: Windows [11 / version]
  
• CPU/GPU: Intel Core Ultra [model] (Lunar Lake, Arc Xe2 iGPU)
  

I installed:

1. Downloaded vsmlrt-windows-x64-cpu.v15.16.7z from https://github.com/AmusementClub/vs-mlrt/releases.
   
2. Downloaded models.v15.16.7z from the same release.
   
3. Extracted the runtime archive into: c:\Program Files\Hybrid\64bit\vsfilters\DenoiseFilter\Vs-mlrt\
   

1. Extracted the models into:  c:\Program Files\Hybrid\64bit\vsfilters\DenoiseFilter\Vs-mlrt\models\
   

Hybrid detects the plugin. In the Filtering → Vapoursynth filter dropdown at the bottom of the window, I can see "DeNoise - DPIRmlrtDenoise". 
Other mlrt-based entries appear too (e.g. Waifu2x NCNN under Filtering → Frame → Resize works fine as a resizer).
On the Filtering → Vapoursynth → DeNoise tab, there is no UI panel for DPIRmlrtDenoise. The tab contains panels for VagueDenoiser, DFTTest, FluxSmooth, etc. — but nothing for DPIR(mlrt).
Selecting "DeNoise - DPIRmlrtDenoise" in the bottom Filter(s): dropdown and clicking + to add it to the chain – nothing happens.

I will be grateful for any help/advice

Thank you

David]]> I'm trying to get DPIR (via vs-mlrt) working in Hybrid so I can denoise on my Intel Lunar Lake iGPU through the OpenVINO backend. The filter shows up in Hybrid's dropdown but I can't actually configure or run it. Hoping you can point me in the right direction.

System:

• OS: Windows [11 / version]
  
• CPU/GPU: Intel Core Ultra [model] (Lunar Lake, Arc Xe2 iGPU)
  

I installed:

1. Downloaded vsmlrt-windows-x64-cpu.v15.16.7z from https://github.com/AmusementClub/vs-mlrt/releases.
   
2. Downloaded models.v15.16.7z from the same release.
   
3. Extracted the runtime archive into: c:\Program Files\Hybrid\64bit\vsfilters\DenoiseFilter\Vs-mlrt\
   

1. Extracted the models into:  c:\Program Files\Hybrid\64bit\vsfilters\DenoiseFilter\Vs-mlrt\models\
   

Hybrid detects the plugin. In the Filtering → Vapoursynth filter dropdown at the bottom of the window, I can see "DeNoise - DPIRmlrtDenoise". 
Other mlrt-based entries appear too (e.g. Waifu2x NCNN under Filtering → Frame → Resize works fine as a resizer).
On the Filtering → Vapoursynth → DeNoise tab, there is no UI panel for DPIRmlrtDenoise. The tab contains panels for VagueDenoiser, DFTTest, FluxSmooth, etc. — but nothing for DPIR(mlrt).
Selecting "DeNoise - DPIRmlrtDenoise" in the bottom Filter(s): dropdown and clicking + to add it to the chain – nothing happens.

I will be grateful for any help/advice

Thank you

David]]>