(06.10.2025, 13:37)Selur Wrote: about Retinex: try installing Microsoft Visual C++ Redistributable 2013 Package
about using HAVC with Retinex like Dan64 showcased: that would require the non-public test version, only Dan64 has access to, and an updated torch environment. (the public releases of HAVC do not support this atm.)
Cu Selur
Thank you, Selur and Dan64
this works,
I'll wait for the stable version with patience
(05.10.2025, 10:36)Selur Wrote: Uploaded a new test version.
Cu Selur
Thanks for the update.
I hope that you can add a change in ColorFix parameter list, as described below:
:param ColorFix: This parameter allows to reduce color noise on specific chroma ranges.
Allowed values are:
'None',
'Retinex/Red' <-- NEW ITEM
'Magenta',
'Magenta/Violet', (default)
'Violet',
'Violet/Red',
'Blue/Magenta',
'Yellow',
'Yellow/Orange',
'Yellow/Green'
I added the new item 'Retinex/Red' with this setting, Retinex will be applied only to DDColor and its family and then will be applied a filter on "red" range of colors (similar to 'Violet/Red').
In my opinion this is the safest way to add Retinex in the coloring pipeline and it represent the 3rd way to add Retinex in HAVC.
I completed the development of HAVC 5.5.2 (attached last RC), apart from the changes already discussed I introduced another enhancement.
Now all the HAVC_() functions are able to accept any clip format, so there is no need to convert to RGB24.
The conversion will be performed internally and the output clip will be in the same format of input clip.
The only exception is if the input clip has format vs.GRAY, in this case the output format will be vs.YUV420P8 (matrix "709") limited range.
Just for curiosity I asked to GPT-5 if is able to colorize videos and because these LLM models are assertive, the answer was yes.
So, after having answered some more question to GPT-5, I provided a short clip to GPT-5 and provided further guidance on the colorization.
When finally GPT-5 was in the condition to colorize the clip, GPT-5 wrote:
However, to be transparent: I can't directly render and recompose video internally at the moment. DeOldify is currently the most effective and accessible solution for automatically colorizing black and white videos with modern, realistic colors, just the way you want them.
So after having invested billions in hardware and software in AI, to colorize the clips the best option remains DeOldify.
As a consequence, since HAVC is an improved extension of DeOldify, it is safe to say that HAVC currently remains the best option for colorizing movies.
These 2 version in my opinion lacks of color stability.
[url=https://archive.org/details/young.-frankenstein.-1974.1080p.-brrip.x-264.-yify-62553139-alq-13][/url]
Dan
These 2 version in my opinion lacks of color stability.
[url=https://archive.org/details/young.-frankenstein.-1974.1080p.-brrip.x-264.-yify-62553139-alq-13][/url]
Dan
Dan64, what you did with Frankenstein is incredible - there is hardly any software that can achieve even a close to similar perfection in coloring.
I myself am trying to colorize an old football match, but so far I am far from your achievements.
# Imports
import vapoursynth as vs
# getting Vapoursynth core
import sys
import os
core = vs.core
# Import scripts folder
scriptPath = 'E:/Hybrid/64bit/vsscripts'
sys.path.insert(0, os.path.abspath(scriptPath))
# loading plugins
core.std.LoadPlugin(path="E:/Hybrid/64bit/vsfilters/ColorFilter/Retinex/Retinex.dll")
core.std.LoadPlugin(path="E:/Hybrid/64bit/vsfilters/Support/libllvmexpr.dll")
core.std.LoadPlugin(path="E:/Hybrid/64bit/vsfilters/MiscFilter/MiscFilters/MiscFilters.dll")
core.std.LoadPlugin(path="E:/Hybrid/64bit/vsfilters/SourceFilter/LSmashSource/LSMASHSource.dll")
# Import scripts
import color
import vsdeoldify as havc
import validate
# Source: 'D:\starlight\Crvena_Zvezda_Fiorentina_1966.mkv'
# Current color space: YUV420P8, bit depth: 8, resolution: 1280x960, frame rate: 25fps, scanorder: progressive, yuv luminance scale: limited, matrix: 709, format: AVC
# Loading D:\starlight\Crvena_Zvezda_Fiorentina_1966.mkv using LWLibavSource
clip = core.lsmas.LWLibavSource(source="D:/starlight/Crvena_Zvezda_Fiorentina_1966.mkv", format="YUV420P8", stream_index=0, cache=0, prefer_hw=0)
frame = clip.get_frame(0)
# setting color matrix to 709.
clip = core.std.SetFrameProps(clip, _Matrix=vs.MATRIX_BT709)
# 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_BT709), if it is not set.
if validate.primariesIsInvalid(clip):
clip = core.std.SetFrameProps(clip=clip, _Primaries=vs.PRIMARIES_BT709)
# setting color range to TV (limited) range.
clip = core.std.SetFrameProps(clip=clip, _ColorRange=vs.RANGE_LIMITED)
# making sure frame rate is set to 25fps
clip = core.std.AssumeFPS(clip=clip, fpsnum=25, fpsden=1)
# making sure the detected scan type is set (detected: progressive)
clip = core.std.SetFrameProps(clip=clip, _FieldBased=vs.FIELD_PROGRESSIVE) # progressive
# changing range from limited to full range for
clip = core.resize.Bicubic(clip, format=vs.YUV420P8, range_in_s="limited", range_s="full")
# setting color range to PC (full) range.
clip = core.std.SetFrameProps(clip=clip, _ColorRange=vs.RANGE_FULL)
# adjusting color space from YUV420P8 to RGB24 for vsHAVC
clip = core.resize.Bicubic(clip=clip, format=vs.RGB24, matrix_in_s="709", range_in_s="full", range_s="full")
# adding colors using HAVC
clip = havc.HAVC_main(clip=clip, Preset="medium", CombMethod="Constrained-Chroma", VideoTune="morevivid", ColorFix="none", ColorTune="none", BlackWhiteTune="none")
# stabilize colors using ColorAdjust (HAVC)
clip = havc.HAVC_ColorAdjust(clip=clip, BlackWhiteBlend="False", Strength=4, ScThreshold=0.09)
# 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=235, gamma=0.90)
# adjusting color space from RGB24 to YUV444PS for vsTweak
clip = core.resize.Bicubic(clip=clip, format=vs.YUV444PS, matrix_s="709", range_in_s="full", range_s="full") # additional resize to allow target color sampling
# adjusting color using Tweak
clip = color.Tweak(clip=clip, hue=0.00, sat=0.95, cont=0.90, bright=-0.00392157, coring=True)
clip = core.std.Crop(clip=clip, left=2, right=2, top=0, bottom=0)# cropping to 1276x960
# adjusting color space from YUV444PS to YUV444P10 for vsRetinex
clip = core.resize.Bicubic(clip=clip, format=vs.YUV444P10, dither_type="error_diffusion")
# color adjustment using Retinex
clip = core.retinex.MSRCP(input=clip, sigma=[25,80,250], fulls=True, fulld=True)
# adjusting output color from: YUV444P10 to YUV420P10 for x265Model
clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P10, range_in_s="full")
# set output frame rate to 25fps (progressive)
clip = core.std.AssumeFPS(clip=clip, fpsnum=25, fpsden=1)
# output
clip.set_output()
Could you take a look, if you have time at all, and give some advice on how to improve the image? I am also bothered by the green halo around the players, I don't know what it is due to.
(08.10.2025, 18:47)Selur Wrote: @Dan64: Side question: Have you looked into using diffusion based models to colorize images to use them as reference ?
In the past I installed StabeDiffusion, but the main tools available are to generate or improve the images based on prompts. To colorize the only tool available in StableDiffusion is DeOldify. Moreover the current version of StableDiffusion is not supporting the RTX50 and it is necessary install the dev version which has a limited support of tools.
But the problem is not to generate the colors but to keep them consistent across the frames.
(08.10.2025, 19:44)didris Wrote: Dan64, what you did with Frankenstein is incredible - there is hardly any software that can achieve even a close to similar perfection in coloring.
I myself am trying to colorize an old football match, but so far I am far from your achievements.
Could you take a look, if you have time at all, and give some advice on how to improve the image? I am also bothered by the green halo around the players, I don't know what it is due to.
To me the clip is already colored, even if the colors were almost removed.
You can try to recover them with this script (please install last RC5)
# Imports
import vapoursynth as vs
# getting Vapoursynth core
import sys
import os
core = vs.core
# Import scripts folder
scriptPath = 'D:/Programs/Hybrid/64bit/vsscripts'
sys.path.insert(0, os.path.abspath(scriptPath))
# loading plugins
core.std.LoadPlugin(path="D:/Programs/Hybrid/64bit/vsfilters/Support/fmtconv.dll")
core.std.LoadPlugin(path="D:/Programs/Hybrid/64bit/vsfilters/Support/libllvmexpr.dll")
core.std.LoadPlugin(path="D:/Programs/Hybrid/64bit/vsfilters/MiscFilter/MiscFilters/MiscFilters.dll")
core.std.LoadPlugin(path="D:/Programs/Hybrid/64bit/vsfilters/SourceFilter/LSmashSource/LSMASHSource.dll")
# Import scripts
import color
import vsdeoldify as havc
import validate
clip = core.lsmas.LWLibavSource(source="Crvena_Zvezda_Fiorentina_1966.mkv", format="YUV420P8", stream_index=0, cache=0, fpsnum=25, prefer_hw=0)
clip = clip[20:] # skip test color
clip = core.std.Crop(clip=clip, left=20, right=20, top=4, bottom=4)# cropping to 1240x952
# Resizing using bicubic spline
clip = core.resize.Spline36(clip, width=1280, height=960)
frame = clip.get_frame(0)
# setting color matrix to 709.
clip = core.std.SetFrameProps(clip, _Matrix=vs.MATRIX_BT709)
# 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_BT709), if it is not set.
if validate.primariesIsInvalid(clip):
clip = core.std.SetFrameProps(clip=clip, _Primaries=vs.PRIMARIES_BT709)
# setting color range to TV (limited) range.
clip = core.std.SetFrameProps(clip=clip, _ColorRange=vs.RANGE_LIMITED)
# making sure frame rate is set to 25fps
clip = core.std.AssumeFPS(clip=clip, fpsnum=25, fpsden=1)
# making sure the detected scan type is set (detected: progressive)
clip = core.std.SetFrameProps(clip=clip, _FieldBased=vs.FIELD_PROGRESSIVE)# progressive
# adjusting color space from YUV420P8 to RGB24 for vsColorAdjust
clip = core.resize.Bicubic(clip=clip, format=vs.RGB24, matrix_in_s="709", range_in_s="limited", range_s="full")
# stabilize colors using ColorAdjust (HAVC)
clip = havc.HAVC_ColorAdjust(clip=clip, BlackWhiteTune="strong", BlackWhiteMode=5, BlackWhiteBlend="True", ReColor=False)
clip = havc.HAVC_tweak(clip, hue = 20, sat = 3.5, bright = 0, cont = 1.0, gamma = 1.0)
clip = havc.HAVC_adjust_rgb(clip, strength=0.6, factor=(0.8, 1.10,1))
clip = havc.HAVC_bw_tune(clip)
# adjusting output color from: YUV444PS to YUV420P10 for x265Model
clip = core.resize.Bicubic(clip=clip, format=vs.YUV420P10, matrix_s="709", dither_type="error_diffusion")
# set output frame rate to 25fps (progressive)
clip = core.std.AssumeFPS(clip=clip, fpsnum=25, fpsden=1)
# output
clip.set_output()
Let me know if instead you prefer recolor the clip.
