Dolby Vision Tonemapping?

[Luke 42077]

4 minutes ago, generiq said:

No need to be rude.

Actually I think it was a serious question because sometimes his posts are too long even for me to make it to the end 

[Guest]

41 minutes ago, generiq said:

 Most of us would use ICCs in mpv because they are used for the rendering.

 

39 minutes ago, softworkz said:

This is nonsense, sorry.

Applying an ICC only in mpv, we can see that it is being used for color mapping. This is in a small window and not full screen.

[softworkz 5065]

1 hour ago, Luke said:

1 hour ago, generiq said:

No need to be rude.

Actually I think it was a serious question because sometimes his posts are too long even for me to make it to the end 

It was meant seriously indeed because the post was just two very vague sentences that didn't reference anything I had written and didn't contribute anything valuable to the specific subject.

[softworkz 5065]

45 minutes ago, generiq said:

Applying an ICC only in mpv, we can see that it is being used for color mapping. This is in a small window and not full screen.

This topic is SPECIFICALLY about MPV using the Windows-configured color profile for the monitor via the icc-profile-auto option.

This is not about something where mpv does something with some kind of ICC profile for some kind of purpose.

[Guest]

44 minutes ago, softworkz said:

This topic is SPECIFICALLY about MPV using the Windows-configured color profile for the monitor via the icc-profile-auto option.

This is not about something where mpv does something with some kind of ICC profile for some kind of purpose.

OK, so setting that same ICC as system profile

 

Using icc-profile-auto, the result is the same

[   0.000][v][cplayer] Setting option 'icc-profile-auto' = '' (flags = 4)

 

 

[   0.333][v][vo/gpu-next] Querying ICC profile...
[   0.334][d][vo/gpu-next/libplacebo] Opening new ICC profile
[   0.334][v][vo/gpu-next/libplacebo] Opened ICC profile:
[   0.334][d][vo/gpu-next/libplacebo]     Raw white point: X=76.04 Y=80.00 Z=87.12 cd/m^2
[   0.334][d][vo/gpu-next/libplacebo]     Raw black point: X=0.000000% Y=0.000000% Z=0.000000%
[   0.334][v][vo/gpu-next/libplacebo]     Contrast = 203 cd/m^2 : 0.001 mcd/m^2 ≈ 203000000 : 1
[   0.334][v][vo/gpu-next/libplacebo]     Detected primaries: ITU-R Rec. BT.709 (HD), also sRGB
[   0.334][v][vo/gpu-next/libplacebo]     Transfer function: IEC 61966-2-4 sRGB (CRT emulation)

Video info

[   0.064][i][cplayer] ● Video  --vid=1               (mpeg2video 720x576 25 fps)
[   0.064][i][cplayer] ● Audio  --aid=1  --alang=eng  (mp2 2ch 48000 Hz 192 kbps)

     

 

 

Edited January 3, 2025 by generiq

[Guest]

But if I use the icc directly, the log entry is a little different

[   0.338][d][vo/gpu-next/libplacebo] Opening new ICC profile
[   0.339][v][vo/gpu-next/libplacebo] Opened ICC profile:
[   0.339][d][vo/gpu-next/libplacebo]     Raw black point: X=0.000000% Y=0.000000% Z=0.000000%
[   0.339][v][vo/gpu-next/libplacebo]     Contrast = 203 cd/m^2 : 0.001 mcd/m^2 ≈ 203000000 : 1
[   0.339][v][vo/gpu-next/libplacebo]     Detected primaries: ITU-R Rec. BT.709 (HD), also sRGB
[   0.339][v][vo/gpu-next/libplacebo]     Transfer function: Pure power gamma 2.4

I don't know what significance that has. Other than the missing white point and the change of transfer function. 

[softworkz 5065]

LOL, you are setting the profile that mpv uses anyway as the monitor profile in order to prove - what exactly actually?

Set this profile as your monitor profile: 

CalibratedDisplayProfile-2.icc

[softworkz 5065]

The simple question then: does it make a difference whether you enable/disable icc-profile-auto and mpv is not full-screen?

[softworkz 5065]

18 minutes ago, generiq said:

But if I use the icc directly, the log entry is a little different

Which command lines did you actually use when you tried with and without icc-profile-auto ?

Edited January 3, 2025 by softworkz

[softworkz 5065]

3 minutes ago, softworkz said:

Which command lines did you actually use when you tried with and without icc-profile-auto ?

There you specified your ITU---.icc file, right?

The outcome (mpv using the profile) 's no surprise of course, because when you force it to do that then it does that. But this has nothing to do with icc-profile-auto - it is meant to use the profile that is configured in Windows without specifying any profile in the command line.

What you are doing makes no sense. First you force mpv to use an icc profile from a file. Then you are setting that same file as the monitor profile on Windows, and then you say "look, they are using the same profile.."

Aaaaargggggghjhhhhh!

[Guest]

1 minute ago, softworkz said:

There you specified your ITU---.icc file, right?

The outcome (mpv using the profile) 's no surprise of course, because when you force it to do that then it does that. But this has nothing to do with icc-profile-auto - it is meant to use the profile that is configured in Windows without specifying any profile in the command line.!

The first test, I didn't set that ICC as the system profile and just pointed at the directory location. So it was used directly --icc-profile=

Then I set it to system profile and changed to 'auto'. 

I was thinking that applying a profile as system default, that it would be used with auto. But thinking about what you're saying, it might have actually used a different profile that Windows has supplied. I'll have to take a closer look at the log later. That would account for the discrepancy. 

15 minutes ago, softworkz said:

Aaaaargggggghjhhhhh!

I'm not trying to make you crazy

My general point was that whatever profile mpv sees, it uses to colormap.

I'll try to remember to test again tomorrow, when I'm more awake.

[softworkz 5065]

5 minutes ago, generiq said:

I'm not trying to make you crazy

But you did well  

[softworkz 5065]

6 minutes ago, generiq said:

My general point was that whatever profile mpv sees, it uses to colormap.

"Sees?`" just don't specify any icc stuff. It's wrong. You don't apply profiles for playback on a desktop (windowed mode). . Monitor color correction is the job of the OS. You have no clue about the subject - and you don't have to. Don't specify any icc stuff on the command line and mpv will do things right without your "help".

In exclusive full-screen mode, you can specify icc-profile-auto, so that mpv uses the same monitor profile like Windows once it takes over control of the output. That's all you need to care about. Getting some sleep is a good idea for sure  

[softworkz 5065]

(other operating systems don't have color management like Windows, and that's where MPV has functionality on board which are not available as platform features.

[Guest]

I only used an icc when I would calibrate my display. I don't need to to that anymore. 

Your point is well taken.

[JoeAverage92 6]

11 hours ago, softworkz said:

 

This doesn't change when I move the Emby app window around on to one or the other monitor, make it full-screen or windowed - always the same: Windows applies the monitor profile to the video.

Please note that mpv can never get into exclusive full-screen mode in the Emby app.

Now, with that setup I have set icc-profile-auto to yes and no and it makes absolutely no difference - just as expected, because it is not on mpv to do any color management in this setup. Windows does that.

What's happening on your side - I cannot say, but maybe this information can help you find out..

First I apologise because I forgot to thank the Emby Team for the new app. Excellent work!

I have done your test using icc profiles that are visibly off (here) and I confirm that mpv (without the flag icc-profile-auto) and emby are both color managed and produce the same results. What is troubling is that I would expect to see a red rec709 less saturated than a red bt2020 (like here) in Emby (my ICC profile was generated by displaycal using Display Pro HL). 

So my guess is that maybe icc-profile-auto in mpv uses the display profile to do some kind of color conversion to display correctly the red rec709 versus red bt2020 (like a 3DLut in madVR) or maybe it is broken and should not be used.

Anyway, thank you again for your replies and sorry if it is not the right post to discuss that.

 

[softworkz 5065]

9 hours ago, JoeAverage92 said:

I have done your test using icc profiles that are visibly off (here) and I confirm that mpv (without the flag icc-profile-auto) and emby are both color managed and produce the same results.

When you run mpv in windowed mode (not fullscreen, not maximized), where it can overlap with other windows and can be before or behind them - do you see a difference in the output with and without the flag icc-profile-auto?

[JoeAverage92 6]

The image remains the same as I move the window over or behind other windows (red is always less saturated with the flag wherever I put the mpv windows). 

Edited January 3, 2025 by JoeAverage92

[softworkz 5065]

2 hours ago, JoeAverage92 said:

The image remains the same as I move the window over or behind other windows (red is always less saturated with the flag wherever I put the mpv windows). 

I meant comparing with and without setting the flag icc-profile-auto - in that situation.

[Guest]

10 minutes ago, softworkz said:

I meant comparing with and without setting the flag icc-profile-auto - in that situation.

So you're trying to determine if the image render is different whether icc-profile-auto is enabled or not? I imagine the difference would be subtle, wouldn't it?

Edited January 4, 2025 by generiq

[softworkz 5065]

36 minutes ago, generiq said:

So you're trying to determine if the image render is different whether icc-profile-auto is enabled or not? I imagine the difference would be subtle, wouldn't it?

You should really read my post...

[JoeAverage92 6]

I have done the test with and without the flag, moving the windows everywhere. I don't see anything changing (without the flag Rec709 Red is too saturated).

I have enclosed the files that I am using. On my calibrated wide gamut screen there is a notable difference between Rec709 and BT2020 (more saturated) with the flag active in MPV. Without, no difference, like in Emby. 

On the screenshot, with the flag, I can see that MPV is doing something based on the windows profile (there's ICC 3DLUT written multiple times in the log).

I don't know exactly what is going on and my configuration maybe broken.

However, it seems to me that there should be a difference (on a calibrated wide gamut screen) between both files when I play them in Emby. Maybe it could be done via an advanced option in Emby that would allow to use a 3DLut (MPV can do this) if the flag icc-profile-auto is unclear.

[Guest]

Something to consider here is that mpv by default will load embedded profiles. Looking at that site you linked,  it appears that their samples have embedded profiles. Can you confirm if your test samples do or do not have embedded profiles?

[Guest]

On 1/2/2025 at 8:01 PM, softworkz said:

Set this profile as your monitor profile: 

CalibratedDisplayProfile-2.icc 9.94 kB · 2 downloads

I just noticed that profiles only get added to SDR, because...

This would explain why I saw no changes.

It looks like ASUS has calibrated my display for HDR. ASUS made an ICC for SDR only. So I can't actually test this, because I always run in HDR. 

mpv with fullscreen exclusivity changes nothing.

[Guest]

And this appears to be a good test page...

https://www.wide-gamut.com/test