Why does mpv report display transfer as "srgb" when using an icc profile calibrated to "bt.1886"?

[OMGImgreeen]

I calibrated my main monitor with displaycal to bt.1886 TRC. I pass the ICC profile to mpv by using icc-profile="path/to/bt.1886/profile". But while playing media the reported transfer function in the display section is "srgb" (while reproducing a bt.1886 media).

[kasper93]

Target transfer function is estimated from the ICC profile as best guess given the gamut that ICC profile maps to. It should always be bigger than the ICC profile itself, but depending on the display response the "guess" may be not accurate.

"srgb' should rather not happen often, which version of mpv do you use, there was small window where icc profile were always mapped to default transfer with colorspace hint, but this shouldn't affect image in most cases.

[OMGImgreeen]

I'm not very familiar with the topic, but thanks for the answer.
From what I understood, mpv tries to estimate the TRC function from the ICC profile and makes a 'best guess,' and that is what I see in the display section of the OSD. Even if it doesn't match the tone curve I used during calibration, that doesn't mean the correction isn't working.
Also, I would like to understand whether the 3DLUT that mpv builds from the ICC profiles already includes the corrections applied by the VCGT table loaded into the GPU.

[kasper93]

From what I understood, mpv tries to estimate the TRC function from the ICC profile and makes a 'best guess,' and that is what I see in the display section of the OSD. Even if it doesn't match the tone curve I used during calibration, that doesn't mean the correction isn't working.

Yes, but in case you can compare results with --target-colorspace-hint=no, not all platforms support client side ICC profile. You compositor may be doing conversions. Need to validate with test pattern in mpv itself.

Also, I would like to understand whether the 3DLUT that mpv builds from the ICC profiles already includes the corrections applied by the VCGT table loaded into the GPU.

3DLUT is built from ICC profile alone. Whether there is correction after that is not considered by mpv.

[OMGImgreeen]

Yes, but in case you can compare results with --target-colorspace-hint=no, not all platforms support client side ICC profile. You compositor may be doing conversions. Need to validate with test pattern in mpv itself.

Not quite understanding here, doesn't -target-colorspaec-hint only affect colorspace? At least that is what I understand from the documentation.

3DLUT is built from ICC profile alone. Whether there is correction after that is not considered by mpv.

Understood, so I should actually disable the VCGT corrections that the OS is loading to the GPU to avoid double gamma correction by mpv.

[kasper93]

Not quite understanding here, doesn't -target-colorspaec-hint only affect colorspace? At least that is what I understand from the documentation.

What this option does is notifying the rendering api what colorspace is the video so it can possibly be converted if needed to target colorspace. It may be possible that this causes unwanted conversion.

[OMGImgreeen]

Not quite understanding here, doesn't -target-colorspaec-hint only affect colorspace? At least that is what I understand from the documentation.

What this option does is notifying the rendering api what colorspace is the video so it can possibly be converted if needed to target colorspace. It may be possible that this causes unwanted conversion.

I've just tried binding 'target-colorspace-hint' to a keyboard button and cycling its value, but nothing changes in the image or in mpv's OSD. So I can assume the color management is working correcty?