Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Hendrik]

Follow up to the thread from MC31: https://yabb.jriver.com/interact/index.php/topic,136378.0.html

After the vacation period and some other work that took precedence around launching MC32, I'm back working on video topics and JRVR.

The last state that I'm working on is that we're reasonably happy with colors now, but the dynamic tone mapping (eg. peak detection) needs some work to produce a more stable brightness experience.

I have collected the 10 samples from Mannis test set (documented here: https://yabb.jriver.com/interact/index.php/topic,136378.msg954597.html#msg954597), and will start evaluating changes to the parameters and/or the algorithm itself to avoid these brightness changes where possible.

Watch this space for updates as we move forward.

[Manni]

Great, thanks!

[jmone]

Sounds good!

[Smack]

Keep up the good work.

[thorsten]

Nice! Hope you had a good vacation

[lello]

Very interesting, thank 

[zybex]

Thanks Hendrik!

Lets hope this thread doesn't morph into a 900+ forum page discussion like the MadVR tonemapping beta thread in AVS Forum

The first post contains lots of interesting snippets, including a description from MadShi about how MadVR does tonemapping. MadVR betas (posted in that thread) have lots of adjustable tweaks and knobs to help development, but I find it overkill. I hope that's not needed here.

I see Manni is heavily involved in that thread too

[Bob Sorel]

Question: Is this discussion limited to the tone mapping aspect of JRVR, or will other aspects of the VR, like subtitle placement, black bar detection, auto AR control, support for CIH users, and other features be allowed?

[Hendrik]

This discussion is specifically about tone mapping. There can be more then one thread, however, forums usually have many of them!

[Bob Sorel]

Thanks Hendrik...got it!

And thanks for all of your work on JRVR...it looks really great! On topic, my observation on a 158" screen is that the tone mapping looks pretty much flawless at medium to high APLs, and that I see minor brightness fluctutation only at low to very low APL levels, and those low APL levels also seem to pick up additional grain that is not apparent at higher APLs. I do not know whether the tone mapping is responsible for the addtional grain or not, but I will add that on smaller screens (like around 85") the grain is not so visible (but stll there), though I can still see the minor brightness flashing.

[tixi]

Hi

I will update to MC32 this week.

I've made some movies test with my setup https://yabb.jriver.com/interact/index.php/topic,138243.0.html and I could test Manni's samples with it if you want.

[Hendrik]

I've been tuning the peak detection parameters to be more conservative and avoid/reduce sudden brightness changes. Although for a full fix we would want some algorithm changes to better account for image properties, instead of only basing this on the plain average and peak.

I'm also testing the "pop" preset mattkhan proposed, to see if it makes a good default on a variety of brightness levels, or a variant of it, to give a bit more of a "lively" image at low brightness.

[tixi]

Oh yes, pop preset,  great !
Thanks

[aaronthe barron]

Hi ,

Just checking how can i tune JRVR like madvr for 4070 super card. Like running on best settings.

Thanks

[Hendrik]

Hi ,

Just checking how can i tune JRVR like madvr for 4070 super card. Like running on best settings.

Thanks

For setup help please open a new thread, this one is for discussing the changes to tone mapping inside JRVR.

[Hendrik]

I ended up creating a new "Spline Curve" selector, with Classic/"Pop"/Custom, custom unlocking the Spline Contrast option, as well as Knee Adapt and Knee Default now.
This should hopefully let us easily swap between these options and test them some more.

I have been testing "Pop" on various files and brightness levels, and decided to just make it the default for now, considering we have an easy option to swap back to the old way as well, and it giving an overall more pleasing picture - even if thats subjective.

[Bob Sorel]

Hendrik,

If it is not too much trouble, please post ANY changes to JRVR in the changelogs of new builds so that we know what to expect and what we should be looking for as progress is being made in tonemapping. For example, your last post indicated changes that are in progress, but it would be nice to know which build actually employs the changes in the VR. To be honest, the "next" build (which is often the way things are mentioned in some of the threads) doesn't mean much to those of us who are not beta testers...actual build numbers would be much better.

Thanks in advance.

[Hendrik]

All changes are always noted in build changelogs when the build is publicly released.

[Hendrik]

Available in 32.0.25 now -

Changed: Tuned the JRVR Peak Detection parameters to be more conservative to reduce noticeable brightness changes from dynamic tone mapping.
NEW: Introduced a new Tonemapping Curve setting in JRVR Advanced Output selection to pick between different Spline curves for tonemapping.
Changed: The default JRVR tonemapping spline curve has been changed to favor a more lively "popping" image in dark scenes.

[Manni]

Available in 32.0.25 now -

Changed: Tuned the JRVR Peak Detection parameters to be more conservative to reduce noticeable brightness changes from dynamic tone mapping.
NEW: Introduced a new Tonemapping Curve setting in JRVR Advanced Output selection to pick between different Spline curves for tonemapping.
Changed: The default JRVR tonemapping spline curve has been changed to favor a more lively "popping" image in dark scenes.

Sounds great, thanks. I'm currently projector-less, so I'll try to take a look on a 4K monitor when I have a chance, but it might take a little while before I get back to you.

[Bob Sorel]

My initial impression watching my first movie tonight using build 25 is that the new "pop" preset subjectively looks great! The colors in both low and high APL scenes really did seem to pop, and I did not detect any brightness flashing or other brightness anomalies. The dark scenes looked excellent. Of course this was just a sample of one, and the quality I witnessed could have been inherent in the source. As soon as I get a chance, I will try to compare various scenes from this film using the different spline curves available, and then of course, try out a lot more source material, but so far so good.

[jmone]

Only done tests so far tone mapping to 1,000nits to an OLED.  Looked good on both "real" and "synthetic" test patterns (up to 10,000nits).  Need to run it on the PJ next.

[umberto5585]

Fairly new to the forums and JRiver but have been immensely enjoying the program for my HTPC. I tested the new build last night connected to my Epson 505ub utilizing the new "pop" preset.  Looked great and I did not detect any brightness fluctuations demoing a few specific scenes in films like Mad Max and Parasite that definitely did cause the issue before! Seems like a substantial improvement!

[jmone]

Watched DUNE last night on the PJ.  No issues to report.  All looked good, did not see any pulsing etc.

[tixi]

Watched Oppenheimer last night, great picture and no issues.

Just one question : By default, JRiver output Auto REC 709 . 4K HDR is BT2020. 1080p is BT709. How to switch automatically from BT709 when I'm viewing 1080p and BT2020 when I'm viewing 4K HDR ?
I do it manually on JRiver and my projector each time , not frequently 'cause I'm watching a lot a 4K HDR .

I've to create two different zone ?

Thanks

[Hendrik]

Thats not really how these things interact typically, and TVs/projectors won't expect a different set of color primaries to be used depending on what kind of file you play, or what resolution it is.

By default, most TVs or projectors will expect a SDR signal, using BT709, and only expect BT2020 with HDR.
You can usually change the configuration in the display to indicate which colors are being sent - but it is important that both the display and the PC agree in these settings, otherwise its going to be wrong.

The only exception here is using HDR pass-through, because that conveys active metadata to the display to indicate that its using HDR+BT2020 now, and it should interpret the image in kind.

If you want to use BT709 with 1080p SDR, and BT2020 with 4K, then you can setup profiles in JRVR to do that - but know that you likely also need to change the setting in the display anytime this happens, so its never automatic.

[Bob Sorel]

Last night I tested several of the clips that Manni gathered over in the AVS thread. Here are my observations using build 25, "pop" preset, HDR to SDR mapping on a 158" projector/screen:

1. Wednesday S01E08 [flickering (special case)].mkv - major brightness change as camera pans into open room. I would not have noticed this in normal viewing, but when specifically looking for brightness problems, I was able to spot it, and it was repeatable.
2. LOTR-Flickering Sample.mkv - PASSED - I could not find any problems at all with this clip.
3. Flickering problematic_clip.mkv - PASSED - I could not find any problems with this clip.
4. Dune Part One 1-002.mkv - single brightness change as men ran from darkness into the light. Again, would not have noticed this in normal viewing.
5. Man of Steel 2013 2160p Remux HEVC TrueHD 7.12.mkv - PASSED - I could not find any problems with this clip.
6. 04 MI6 Fallout (Chapter 4) Brightness Stability-002.mkv - PASSED - I could not find any problems with this clip.
7. 03 MI6 Fallout (Chapter 4) Brightness Stability-001.mkv - FAILED - major brightness flashing in ballroom which continued to a lesser extent going into the hall. This was by far the most obvious case of brightness instability I saw, as it was really in my face and could not be overlooked. Funny thing is, I watched the same clip on a couple of flat screens using hardware players (not MC) and the problems existed there as well, so are ALL players having a problem with this scene or is it intentional, as the scene is a disco ballroom type of environment?
8. Jumanji The Next Level (2019) [unstable brightness].mkv - PASSED - I could not find any problems with this clip.
9. 01 Mad Max Fury Road (Chapter 3) - Brightness Stability.mkv - PASSED - I could not find any problems with this clip.

So, for me, I found 1 massive FAIL, 2 very minor FAILS, and 6 PASSES that I could find no problem with whatsoever. I am not the best person to examine these clips, though, as I am nowhere near as astute with my observations as others, so take my observations with a grain of salt. The only clip that I would have noticed in normal viewing is the MI6 ballroom clip, as that was very obvious, but then again, it being shot in a ballroom with disco lights I would not have questioned the brightness flashing.

Now, that being said, the worst case of brightness instability/flashing that I have run across ANYWHERE is pretty much the entire film Love In Taipei (my wife's film, not my thing), as the flashing instability was consistent throughout the film (starting after about 15 minutes) when I watched it the first time using older31/newer32 builds. So I held my breath and watched as much of it again as I could stand watching using build 25 and it was virtually perfect with no anomalies whatsoever!

[JimH]

Thanks for the testing and the excellent report, Bob.

[SamuriHL]

I didn't realize there's a new thread on this for MC32 so that's exciting.    I watched Alien last night on my OLED tone mapped to 900 nits.  Given that Alien is mastered at 1000 nits, it's not doing a whole lot of tone mapping. And Alien does have HDR10+ metadata to help.  I didn't notice any issues at all.  Aliens is waiting to be delivered today so I'll be playing that tonight.  No idea what that's mastered to.  I'm almost at the end of a Defenders marathon so I'll have more time for movies soon and plan on checking this out further.  I love seeing improvements!

[lello]

I'm not sure if I'm in the right thread, but I'll try.
I'm watching a TV series in HDR10+ and during outdoor scenes during the day in naturally bright areas, the brightness is excessive.
Initially I thought it was a feature of the film intended by the director, but the images were truly unwatchable and by doing various tests, I noticed that by removing the checkbox on "use HDR10+ tone mapping curve, if available" the situation improves but not by much: there are other settings I need to change?

For example, I left the Peak Nits target at 103 by default, even though my projector reaches a maximum of 60 nits precisely to avoid images that are too bright

[Hendrik]

It sounds like the scene was mastered weirdly,  especially if disabling HDR10+ improves the situation.

I'm not sure about specific settings to reign in such a scene without seeing it, you might have to try some.

[lello]

It's not just one scene, but all scenes in daytime situations with full sun. I've only seen 2 episodes of this series, but it's always like this.

I'll make further attempts tomorrow

[lello]

Today I made many attempts, including reinstalling build 24 (prior to the introduction of the new Spline curve), but to get a decent image I have to disable HDR10+.

I saw that in my library there are other TV series with HDR10+ so I tried to watch them again, but none of them had very sunny scenes in broad daylight that would allow me to make a comparison, but I noticed that with HDR10+ enabled, the scenes dark were too dark.

In summary, by enabling HDR10+ scenes with highlights are too bright and scenes with lowlights are too dark.

[Hendrik]

HDR10+ relies on bitstream data for brightness of the scenes. Its just what it comes as. We actually don't have HDR10+ enabled by default, because in most cases our own dynamic tone mapping (eg. peak detection) does a better job.
So in short, you don't really lose anything by turning off HDR10+

[Bob Sorel]

Thanks Hendrik...got it!

And thanks for all of your work on JRVR...it looks really great! On topic, my observation on a 158" screen is that the tone mapping looks pretty much flawless at medium to high APLs, and that I see minor brightness fluctutation only at low to very low APL levels, and those low APL levels also seem to pick up additional grain that is not apparent at higher APLs. I do not know whether the tone mapping is responsible for the addtional grain or not, but I will add that on smaller screens (like around 85") the grain is not so visible (but stll there), though I can still see the minor brightness flashing.

The problem that I mentioned in the above quoted post still exists. Am I the only who sees this minor flashing and added grain?

[Manni]

Sorry it took me so long to test this but I was busy and projector-less. Only got my new display (a Samsung S90C 77" QD OLED) last week and just got to calibrate it. I set it to 108nits peak brightness in SDR to a have similar evironment as what I had when I tested brightness stability with V31 on my JVC NZ8 (now sold).

The latest JRiver V32 with JRVR passed all the tests in my list, except the last one (Dune). So congrats to Hendrik and the team, it's a huge improvement.

Details: https://www.avsforum.com/threads/brightness-stability-issue-with-projectors-or-low-nits-displays-when-using-dtm-with-jrvr-and-madvr-no-support-questions.3290169/#post-63007548

Please note that as I don't use DTM anymore but HDR passthrough and static tonemapping on my new display (1,150 nits peak brightness) I won't spend much time working on DTM.

I'm also using JRVR rather than madVR as I don't need the vertical picture shift anymore, thanks to QD OLED blacks.

Thanks for all the good work!

[Manni]

Last night I tested several of the clips that Manni gathered over in the AVS thread. Here are my observations using build 25, "pop" preset, HDR to SDR mapping on a 158" projector/screen:

1. Wednesday S01E08 [flickering (special case)].mkv - major brightness change as camera pans into open room. I would not have noticed this in normal viewing, but when specifically looking for brightness problems, I was able to spot it, and it was repeatable.
2. LOTR-Flickering Sample.mkv - PASSED - I could not find any problems at all with this clip.
3. Flickering problematic_clip.mkv - PASSED - I could not find any problems with this clip.
4. Dune Part One 1-002.mkv - single brightness change as men ran from darkness into the light. Again, would not have noticed this in normal viewing.
5. Man of Steel 2013 2160p Remux HEVC TrueHD 7.12.mkv - PASSED - I could not find any problems with this clip.
6. 04 MI6 Fallout (Chapter 4) Brightness Stability-002.mkv - PASSED - I could not find any problems with this clip.
7. 03 MI6 Fallout (Chapter 4) Brightness Stability-001.mkv - FAILED - major brightness flashing in ballroom which continued to a lesser extent going into the hall. This was by far the most obvious case of brightness instability I saw, as it was really in my face and could not be overlooked. Funny thing is, I watched the same clip on a couple of flat screens using hardware players (not MC) and the problems existed there as well, so are ALL players having a problem with this scene or is it intentional, as the scene is a disco ballroom type of environment?
8. Jumanji The Next Level (2019) [unstable brightness].mkv - PASSED - I could not find any problems with this clip.
9. 01 Mad Max Fury Road (Chapter 3) - Brightness Stability.mkv - PASSED - I could not find any problems with this clip.

So, for me, I found 1 massive FAIL, 2 very minor FAILS, and 6 PASSES that I could find no problem with whatsoever. I am not the best person to examine these clips, though, as I am nowhere near as astute with my observations as others, so take my observations with a grain of salt. The only clip that I would have noticed in normal viewing is the MI6 ballroom clip, as that was very obvious, but then again, it being shot in a ballroom with disco lights I would not have questioned the brightness flashing.

Now, that being said, the worst case of brightness instability/flashing that I have run across ANYWHERE is pretty much the entire film Love In Taipei (my wife's film, not my thing), as the flashing instability was consistent throughout the film (starting after about 15 minutes) when I watched it the first time using older31/newer32 builds. So I held my breath and watched as much of it again as I could stand watching using build 25 and it was virtually perfect with no anomalies whatsoever!

Hi Bob,

There is no issue in the ballroom part of the MI6 clip. It's the flashing lights in the scene itself. When there is an issue in that clip, it's when the two agents (Tom and Henry) enter the bathroom and two guys walk out. When they walk out, there can be a brightness jump. That clip is a pass with the latest JRiver V32.

You might want to look at my notes in the screenshots in the thread I linked to, there is usually a short description of what to look for in each clip.

Some of the clip provided are not related to brightness intability but with flicker, which is a separate issue. I've only looked at the clips that I had specifically selected for brightness instability, and JRVR now passes them all but one, at least here.

Thanks for your feedback!

[Bob Sorel]

Hi Bob,

There is no issue in the ballroom part of the MI6 clip. It's the flashing lights in the scene itself. When there is an issue in that clip, it's when the two agents (Tom and Henry) enter the bathroom and two guys walk out. When they walk out, there can be a brightness jump. That clip is a pass with the latest JRiver V32.

Hi Manni....yeah, I thought that might be the case, and now you have confirmed it for me...thanks!

You might want to look at my notes in the screenshots in the thread I linked to, there is usually a short description of what to look for in each clip.

I had downloaded the clips that I tested a few months ago when you first started the thread, and at that time there were no notes on what to look for. I revisited the clips since reading the notes, but the clips I collected still looked fine to me, so I did not re-report them here. I did notice the instability in Dune and reported it as I saw it. I think you now have a second Dune clip in your list, so I will revisit both Dune clips.

Some of the clip provided are not related to brightness intability but with flicker, which is a separate issue. I've only looked at the clips that I had specifically selected for brightness instability, and JRVR now passes them all but one, at least here.

Yes again... I was looking for both instability and flashing, as the flashing is something that I personally find more distracting in normal viewing. I found a worst case scenario in the entire movie Love in Taipei, which was a flash-fest the first time I viewed it with an older version of MC32, but starting with build 25 (the build I used for testing) the flashing was completely gone, and in subsequent viewing of more material, I have not seen that annoying flashing in any movie or video.

Thank you for the clarifications!

Edit: I just revisited the 2 Dune clips and I discovered that the original one I reported on was indeed clip #2. As in my original report, this is what I found originally, and the clip looked the same to me this time also:

4. Dune Part One 1-002.mkv - single brightness change as men ran from darkness into the light. Again, would not have noticed this in normal viewing.

Manni, I beieve this coincides with your observation despite the difference in the way that we described the scene.