HDR metadata not correct, replaced by standard values

The metadata for (as it seems) all HDR10 content played on Infuse is replaced with BT.2020 Mastering Display Color Primaries, a Mastering Display Luminance of 1000/0.0050 and MaxCLL/FALL of 4000/1000. This is not correct. The metadata should reflect the actual source values.

… does this impair the picture quality? ?

Yes, it does. This means that e.g. if you are using custom gamma curves for different MaxCLL values etc. they could be wrong. In simple terms, it influences blacklevels, brightness, contrast etc. I’m sure there are more disadvantages for different screens or setups. But basically, the HDR metadata should not be tampered with by the AppleTV or Infuse at all.

When I convert an UHD-BD to a mkv via MakeMKV, are these metadata correctly extracted and stored in the mkv?

This sounds like it is a serious problem. Maybe James can comment on that. I hope a fix is possible.

You should ask the makers of makeMKV…

From what we can tell, these values are being set by the Apple TV and cannot be adjusted by Infuse.

I will open a ticket with Apple to report this issue, as ideally we should have more control over the metadata that is being sent.


Okay, thank you.

But in what cases is this error of relevance? Only when I manually set custom gamma curves based on the wrongly reported metadata?

@jowi How did you find this out?
@ix400 It’s relevant because this would have the greatest impact on picture quality, and a reasonable explanation why some, but not all, UHD titles look wrong when played through Infuse.

In defense of Infuse, HDR is still a moving target even on the side of the standards bodies that write the spec. For example, there is no clear spec for proper tone mapping, just a set of recommendations with a lot of wiggle room. But I guess with the advent of OLED displays which have permanent color drift from day one the industry stopped caring about technical correctness.

Infuse should not do the tone mapping.

I did not find it out myself; it was noticed by someone else on another forum. So i can’t take credit for it personally, but i did notice that even though i’ve setup macro’s in my HDFury Vertex for specific 1000nits and 4000nits gamma’s, i noticed that only the 4000nits gamma was ever selected, never the 1000nits one, even when i knew from the specs that the movie was 1000nits, which made me suspect that the ATV (or Infuse) was not outputting the HDR metadata correctly.

At the moment the ATV is my only 4K HDR player, so i can’t compare results with another player yet. If someone can, please do so and share the results!

Yes, that would be a good explanation why UHD movies often look wrong in Infuse. For example, Blade Runner UHD has these values from the stream:

Maximum Content Light Level: 1655 cd/m2
Maximum Frame-Average Light Level: 117 cd/m2

Which look like much more sensible values than the generic 4000/1000. Specifically, a generic value of 1000 when the stream is actually at 117 would be a good explanation why the picture looks “washed out” and “too dark”.

It’s my understanding that Infuse 5.6.11 fixes this.

… the question is, what the TV / Projector is doing with this kind of meta data.
I have set my VW760 manually to clip at around 900 nits. Hence, it should not make a difference what the ATV does with the meta data.

Will LG’s Active HDR negate this problem?

Can someone tell me an example, which UHD movie is not played correctly (with respect to HDR), caused by this meta data problem?

I would like to test this with my Oppo as reference.

Did you get any response from Apple?

I don’t know if they are correct or complete but MakeMKV output does have colour space and master display metadata.
I also see that some mkvs have a little more, others a little less metadata which makes me think it does copy the title metadata that’s present.
I haven’t seen any titles (I’ve only looked at a few so far) with the two options talked about in this thread.

If you encode UHD BDs with handbrake only the colour space gets set, while ffmpeg expects you to set this yourself.
Also either the master display or the options discussed here need to be set to have the encoded video trigger as UHD.
I don’t know what the implications are of not specifying the mastering display parameters but I have assumed it is needed as a minimum with the two other options being optional, as required.

No, but Apple rarely ever comments on plans for future API changes before they are released.

Want to add to this. The new JVC projectors are coming soon. They have auto tone mapping features that will key off the maxCLL/FALL values. This is a big selling point on them.

I’ll be upgrading my x790 to an N7. Infuse fixing this as 4000/1000 basically ensures that the JVC algorithm will not work and likely bad imaging will result.

Shame on the APIs. ?

I came across this as the new owner of a JVC RS3000 and was seeing the 4000/1000 MaxCLL/FALL values passed from both infuse and Plex’s player so yes you are correct this makes the JVCs Auto Tone Mapping function worthless. Interestingly enough Netflix is able to pass the correct MaxCLL/FALL value through so I’m not sure if Apple has given them a private API or what.

Has there been any update on this issue? I have a JVC NX7 and seeing the “standard” metadata coming though – makes HDR content practically unviewable.