Mpv: Windows 10bit output & Infinite contrast ICC Profile

[sat4all1]

mpv version and platform

mpv-x86_64-20180701-git-861c102
Windows 10 x64 1803

Reproduction steps

• 10/12bit output issue:
  Nvidia control panel: 12bit for 23...30hz
  using autospeedwin with mpv to match the video refresh rate.
• ICC Profile issue:
  My LG OLED is set to 2.4 gamma, wide color gamut, 100nit

Expected behavior

• mpv output 10 or 12bit
• mpv use bt.1886 2.4 gamma when using my oled profile.
  I'm using the same profile to create 3dlut for other player and it works perfect.

Actual behavior

• mpv still report 8bit as display bitdepth
• mpv detect infinite contraste and fall to 1000, which make the picture look washed out like using 1.9 gamma.
  I've tried icc-contrast=100000 and 1000000, but it crashes blacks and look very dim.

Log file

https://pastebin.com/SACVbdq7

[zc62]

10 bit issue, see #5237 and #5804

[sat4all1]

If i understand correctly, 10bit output isn't supported yet?
About the icc profile issue, i think i found the sweet spot for my display "icc-contrast=9969", but there are still problems. when using icc profile i get pinkish tint
mpv icc profile on

mpv icc profile off

madvr 3dlut on "The 3d-lut was created from the same profile i'm using with mpv"

here is the icc profile
my lg oled.zip

log.txt

[haasn]

If i understand correctly, 10bit output isn't supported yet?

It's supported on Linux, but apparently not (yet?) on Windows or OS X (?).

About the icc profile issue, i think i found the sweet spot for my display "icc-contrast=9969", but there are still problems. when using icc profile i get pinkish tint

Can you upload the profile?

[sat4all1]

@haasn

here
my lg oled.zip

thanks

[Akemi]

It's supported on Linux, but apparently not (yet?) on ... OS X (?).

there is only an open PR (#5623) for 10bit support on macOS.

[sat4all1]

@haasn

any hint?

[sat4all1]

@haasn

Is it possible to allow infinite contrast because even with icc-contrast set to 1000000 i lose perfect blacks.

[zc62]

Made a commit in my fork: https://github.com/zc62/mpv/commit/96c3f8a16cc60b75d6a370d765a2af6560c7ed04. Use --icc-contrast=inf to bypass the code that modifies the black point.

Is this what you really want? Could you test how this is compared to madVR + 3dlut (which software did you use to generate that 3dlut from the profile?), and can you really tell the difference between --icc-contrast=inf and --icc-contrast=1000000?

[sat4all1]

Anywhere to download windows build based on your fork?

[zc62]

I wrote and compiled the code in Linux. It is a bit tedious to make a fully static build to let others directly use the compiled binary, so I am expecting you to follow the guide (https://github.com/mpv-player/mpv/blob/master/DOCS/compile-windows.md) to compile it on your own. It is actually not that hard to compile mpv in MSYS2. Then you can directly run mpv from the MSYS2 shell.

What you need to do to compile that branch is just replacing

git clone https://github.com/mpv-player/mpv.git && cd mpv

by

git clone https://github.com/zc62/mpv.git && cd mpv && git checkout inf_contrast

[sat4all1]

Now with icc-contrast=inf i have perfect blacks, and no visual difference compared to madvr + 3d-lut generated using displaycal.
Hope you can push this to master branch.

Thanks much

[zc62]

Well, I was actually thinking inf should not make too much difference from 1000000. Probably that's because I don't own an OLED TV. Will make it a PR.