I make public. I think anything more and they start to look like something you’d cut out of a fallopian tube.
2 Likes
My hasty color correction done while I had my windows anti blue light settings on. I pull cuz it’s wrong tint as result. Won’t redo, perfs too perfunctory and I too lazy to bother with stuff richter probably learned on the toilet that day.
Now that I look at it as a whole there are surprisingly few legendary piano vids especially on DVD. Damn!
WOH! Insane discovery!
After a week of being very confused, I realized in spite of color space tagging, most modern players don’t read color space tags. Jriver, VLC and even the input panel of Topaz Video ( which is a software designed to restore those old color space files) do not build a full color managed pipeline for Y′CbCr video. It simply hands your MPEG-2 4:2:0 stream off to DirectShow/DirectX for a Y′CbCr→RGB conversion using the standard Rec.709 matrix and TRC, regardless of any 601 or smpte170m etc. tags you have in the file.
I’m using vapoursynth to do archival color space conversions to 709 primary, matrix and TRC using 10 bit so its lossless and shit looks waay different.
Pretty much any DVD sourced or ripped media gonna look a little fucked color wise without a color space conversion to 709 poss.
I had AI do the math on how fucked all DVD media has been since people started watching that shit on computers.
Here’s a quantitative look at how “blind” 601→709 conversions can skew both colour and brightness:
1. Chromaticity (Primaries) differences
– The red, green and blue primaries shift by up to ~3% in x or y—and blue’s y coordinate moves by ~14%—when you use Rec.709 primaries instead of the original NTSC/SMPTE-170M gamut.
– In practice, that can produce noticeable hue shifts, especially in highly saturated blues.
2. Brightness (EOTF) differences
– Applying a Rec.709 transfer curve (gamma 2.4) instead of the CRT-like Rec.601/EOTF (≈gamma 2.2) alters displayed luminance by:
- ~9% darker at very low luma (Y′=0.1)
- ~4.7% darker at Y′=0.3
- ~2.7% darker at mid-grey (Y′=0.5)
- ~1.4% darker at Y′=0.7
- ~0.4% darker at highlights (Y′=0.9)
Bottom line: without properly respecting your SMPTE-170M/Rec.601 tags, modern players can introduce:
- Hue errors up to ~3% (and blue shifts up to ~14%)
- Brightness errors up to ~9% in shadows and ~3% around midtones
Those percentages explain why DVD-sourced material “looks off” without an explicit colour-space conversion.
WUT DA FUQ~? Conspiracy to make us buy HD shit? Mystery…
1 Like
Absolutely
And each TV manufacturer is different
That’s why most modern stuff is graded right into rec. 709 gamma 2.4 for YouTube deliveries
1 Like
That part of the stack makes sense. I was just shocked I was seeing player read metadata and say we are in 601 space with the right transform and matrix but it’s really running a secret conspiracy in it’s engine to treat everything like it’s rec 709 when actually translating to display.
Anyhoo. I’m uploading a properly color space adjusted version of richters geriatric winter wind from barbican. 
My updated workflow.
- Eat cheeseburger and redbull
- Makemkv rip DVD
- Deinterlace with QTGMC and color space convert using vapoursyth stuff in Hybrid - Export 10 bit lossless FFV1 in MKV
- Color grade if not lazy (use regular monitor settings)
- Export YUV FFV1 10 bit
- Import into Topaz and satanic AI models
- Export and finish in neat video if need temporal stabilization.
- Finito.
2 Likes
Lmao.
Sounds good!
1 Like
yup looks proper now.
1 Like
Woh… just learning about how fucked aspect ratio is in DVDs. Same thing as color space. World so steeped in HD they forget the lost arts of pixel shapes and aspect tags. Add to the hyrid stage convert to square pixels in appropriate aspect ratio.
This hobby is insanely tedious and horrible. HOT!
WTF davinci make Ho a stout short king? Fuck.. I guess I really do gotta square pixel aspect convert this sheyat.
1 Like
*** gangbang
Poss… I’m not overweight, I’m just in the wrong aspect ratio. 
#fattythoughts
I bet now that this in square pixels the maths the diffusion models do work better too.
1 Like
Taking a forensic approach to the horowitz rach 3 restoration this time based on errors on the early transfer from broadcast tape I can pinpoint. I map out likely source and playback chains which led to the 1990s DVD release (bluray was just a garbage restoration of this)
Here what I suspect got fucked:
- Pedestal mix-up – Many US facilities forgot to subtract the 7.5 IRE setup before feeding BT.601, so digital “black” lands at code ≈ 32 instead of 16 → lifts shadows, lowers contrast.
- Cheap composite decoder – If they went composite → YCbCr with a one-chip decoder, the notch/comb filter threw away half the chroma bandwidth; saturation plunges.
- Ageing tape – Chrominance on oxide stock fades faster than luma, so a 20-year-old Quad/Type C already looked desaturated before the transfer engineer even pressed Play.
I research how to fix each of these scientifically and not just eyeball it on my janky monitor.
Pray for ho. Maybe he watchable soon without fake colors.
1 Like
This have zero color grading, just insanely complex chroma channel restoration sheyat, the black level set point fix and topaz upscale.
what it look like before
Dude, am I gonna have to colour calibrate my phone to appreciate dis?
1 Like
Best way to have color accurate view of tape deterioration before they transferred it badly yeah.
1 Like
This the most poeple I seen on here this decade. Damn!
Ho rach three with forensic cleanup finish in a couple days.
1 Like
I love y’all
1 Like
pysop
Randomly this Hyrbid program is AMAZING. All the rando open source scripts you’d ever want in one place and it mostly works.
2 Likes
Going deeper still. A lot of the source tags don’t line up with the media on these old zero production classical piano DVDs. Using a tool called DGindex to analyze.
So you got bad transfers of original media throwing away resolution and color info which was already limited, bad tagging of the encoding, and bad identification of the tagging played back through devices that are largely unable to play back in the right color spaces even when they are flagged.
Damn… technology sucks dicks. Glad we are entrusting the world to people who do this to us.
1 Like
can we finetune model on wikid roque piano dvd ?