[DoVi_Scripts] Multi-Function Scripts for Dolby Vision processing and a lot more...

[Kuler087]

Yeah, my C2 has the same banding and posterization (https://drive.google.com/drive/folders/1DxMiJvf_HHTqxw5AW4WdbrKkCkKrH3JG?usp=drive_link) but it's a least 2 times worse on the G5... If they can bring it to the level of the C2, the G5 would be great.

Did a tone mapping comparison (https://www.youtube.com/watch?v=0FaFGmU5hHo) yesterday, and the G5 is so powerful. The Spears and Munsil 10 000nits clip is even watchable without tone mapping on the G5.
Without tone mapping, the C2 is all clipped.

https://i.ibb.co/bG06Nw4/Mpc-hc64-VSs-Uya-LZ0l.png

[SamuriHL]

Good lord that's impressive.

[SeeMoreDigital]

Good lord that's impressive.Indeed...

The new LG Display 4-stack OLED panels are a game changer. Plus there are versions of this panel in the new Panasonic Z95B and Sony Bravia 8 II.

[SamuriHL]

Indeed...

The new LG Display 4-stack OLED panels are a game changer. Plus there are versions of this panel in the new Panasonic Z95B and Sony Bravia 8 II.

I have a 77" G2 so I'm looking at probably next year or the year after for my next upgrade. If this is the direction we're headed, I'm super excited. Now if they can give more bt.2020 coverage while they're at it, we really would be in an amazing position with these displays! Paired with adding DV metadata to everything and running that on an Ugoos....giddyup!

[Hellboy.]

Thanks, the G5 is a beast and I love it, but HDR10 is broken (https://slow.pics/c/AYrYCbvm)(banding and posterization). It doesn't affect me because I generate DV for everything...

FYI, Cars 3 P8 hybrid cannot be made. The HDR10 BD layer is a 600 nits trim, and the DV master is 4000 nits.

Grade check...
Cars 1: https://slow.pics/c/PGCSvxu9
Cars 2: https://slow.pics/c/79SARikb
Cars 3: https://slow.pics/c/RldFk4g9

With the broken HDR10 is it better to let Ugoos AM6Bplus convert everything to DV?

[Kuler087]

It certainly fixes the issues, but I'm not sure how accurate the rest will be... Better just generate proper CMV4.0 metadata in Resolve (https://www.youtube.com/watch?v=FVSh3oGqfXY) or doviscripts 3-1

another banding/posterization comparison (https://slow.pics/c/ENN0RuWy) ... I think I'll find a lot more scenes like this and make a YouTube video about it.

[en6ads]

Just wanted to know your thoughts on the latest ng corelelec nightly. loads of changes (https://github.com/CoreELEC/xbmc/commits/aml-4.9-21.3), even a few commits from cpm.

For TV-led Dolby Vision on the Ugoos, what does A14 do that this latest nightly can't or gets wrong (besides VS10)? Is level 5 not properly implemented in TV-led nightlies?

Thank you.

[Kuler087]

I dont know but you can test everything that works on A14 but doesnt with the nightlies with these:

L5: https://drive.google.com/drive/folders/1koVQa3hyZwZ8QP-36_rkCBLZEUH69zAV?usp=sharing
Positive lift: https://drive.google.com/drive/folders/1LbuwBDZj5x3i1bdbAntsZlr5CiYJ9KKj?usp=drive_link
unaltered metadata: https://drive.google.com/drive/folders/1FJcuC1Eo8xFS9P1wgzpARlJWcWoliuQ4?usp=drive_link

A14 does everything perfectly, I see no reason to ever update again, and I certainly won't update my ugoos anytime soon, not even if A15 comes out.

[tormento]

Is there some utility that can plot/analyze the created .measurements file?

[Kuler087]

workflow 6-2 can plot it, otherwise, open it in:

DoVi_Scripts\tools\madMeasurementAnalyzer.exe

[tormento]

workflow 6-2 can plot it, otherwise, open it in
I have a stupid spike that ruins the average values:

https://i.ibb.co/LzHbCBB7/Better-Man-HDR-plot.png (https://ibb.co/KjSR9BBv)

Any way to exclude spikes or choose starting and ending frame?

Why the percentages on top right are all 0?

[Kuler087]

Spikes like this are usually caused by your lavfilters settings. Your percentages are 0, probably because you did not install Python? You may have to install both the windows store (https://apps.microsoft.com/detail/9pjpw5ldxlz5?hl=en-US&gl=US) and regular (https://www.python.org/downloads/) versions.

My measurement of this movie (https://ibb.co/7xWwtrfK) (letterbox excluded)

https://i.ibb.co/TMbwbgpC/Better-Man-2024-BD-HDR-plot.png
https://i.ibb.co/PvZpdQVZ/rundll32-tf8-F0-NFb-Zn.png

[tormento]

Spikes like this are usually caused by your lavfilters settings.
That is one of the few movies where I have spikes. If you manually scan it with the FranceBB Tektronik plugin, you will see a sudden spike in luma in the movie too.
you did not install Python?
I have latest stable installed.
You may have to install both the windows store and regular versions.
Why? I can't find the rational in what you are saying.
My measurement of this movie
I have the UHD, you scanned the BD. There is no much sense in scanning a BD where of course you can't have spikes.

EDIT:

It finds and calls Python:

https://i.ibb.co/Df9tDK2h/image.png (https://ibb.co/VYgtWB4R)

[Kuler087]

That is one of the few movies where I have spikes. If you manually scan it with the FranceBB Tektronik plugin, you will see a sudden spike in luma in the movie too.
If it's not your lavfilters settings, then your video source has a glitch.

Why? I can't find the rational in what you are saying.
Some workflows use the Windows Store version.
or you are missing one of the required library:
opencv-python, colour-science, scikit-image, matplotlib, numpy, colour, opencv-python-headless , PyQt5

I have the UHD, you scanned the BD. There is no much sense in scanning a BD where of course you can't have spikes.
what lol? BD mean bluray disc...
Of course, I measured the UHD bluray disc... The 1080p bluray disc plot would be in the 100nits...
not to mention that I already measured practically all the UHD-BD that exist... (https://drive.google.com/drive/u/1/folders/154fBNllwOHL4Lckc7wDV8QKFJwFxnDt-)

[tormento]

If it's not your lavfilters settings, then your video source has a glitch.
For sure, there must be some problems on the italian version. Given that, is there any way to ignore glitches? Today I've "solved" it cutting the video before the glitch but there should be some workaround.
or you are missing one of the required library
Is that the complete python requirements or I need more? I really don't like the idea to install python from the store.
I measured the UHD bluray disc
Sorry, my bad.
I already measured practically all the BD that exist
That's awesome.

[Kuler087]

Is that the complete python requirements
yes

For sure, there must be some problems on the italian version. Given that, is there any way to ignore glitches? Today I've "solved" it cutting the video before the glitch but there should be some workaround.

Open the measurement file in the analyzer and look at the 99% peak values:

https://i.ibb.co/KzpnczJn/mad-Measurement-Analyzer-a-Id16-Ahg-Ky.png

[Kuler087]

Updated tutorial for the P8 hybrid maker/grade check:


https://i.ibb.co/XfB4BTmH/Firefox-i-KWpat-Fv2-A.png (https://www.youtube.com/watch?v=J06dUCePdd8)

[wyup]

Hi, Do I need to overwrite files from the latest toolpack or just the MKV/MP4/TS scripts? Because there are many files and it's tough to select only newer files. I also update some tools on my own.
In Changes section there's only hdr10plus.exe update so I guess it's the only change.

[Kuler087]

yes hdr10plus_tool is the only change so no need to update the rest.

[coopzr]

Will running multiple instances of dovi_tools cause any issues? E.g. One instance generating DV and another instance injecting/plotting/something else?

[Kuler087]

Each workflow creates its own temp folder so it's fine, and I do it all the time.
Just don't use two instances of the same workflow.

[wyup]

what's your experience with the LG G5? Does it have a good and reliable Dolby Vision implementation? I read on Rtings that it had some problem with streaming content...

[Kuler087]

Dolby Vision:
This is the best DV TV I’ve ever owned. It is close to perfection, even with the older and buggy cmv2.9 content.
Tone mapping (https://docs.google.com/spreadsheets/d/15i0a84uiBtWiHZ5CXZZ7wygLFXwYOd84/edit?gid=1289366200#gid=1289366200) kicks in when the metadata reaches 2550 nits, which means most content doesn’t need DV or any tone mapping.

Blacks are raised a bit, but it’s much better than on my C2 or C8. On the G5, it’s not visible at all in real content. There’s a faint glow if you use the pure black test file and stare at the screen for a couple of seconds. If you look closely, you can see that the glow is caused by DV’s stronger dithering noise.
Black clipping/crush is almost perfect. Only cmv2.9 with 4000 nits MDL shows a slight crush, but this is 100% better than my C2/C8.

HDR10:
Now, this is a major problem. Dark scenes with highlights look like straight garbage: tons of posterization and banding. There are also some ugly DSE frozen grains. All of these issues only affect HDR10 and not DV, so it doesn’t affect me since I only watch DV content.
LG should fix this issue and at least bring it to the same level as my C2, which also has these issues but they’re almost unnoticeable. Near blacks are also raised, for example, I can easily see code 65 flashing at normal viewing distance. This should not be the case.
Considering DV was also broken at launch, I don’t understand how LG went ahead and released this TV in this state. I mean, this is unacceptable for a TV that cost me 7k CAD.

SDR:
I haven’t watched any SDR content yet, but the accurate mode (FMM) with OLED Light 80 is WAY too bright, a big 500 nits. Unless you want to get blinded in SDR, you must lower the OLED Light to at least 25. I did not measure which value gives exactly 100 nits, but just by looking, I have mine set to 10.

The remote:
What a joke. Didn’t last more than 5 minutes before I threw it away. I’m using my C2 remote.

Gaming or PC monitor:
This is not the TV for you if you use it as a monitor and sit close to the screen. The diagonal dithering lines look horrible. Sometimes I even see them from 8 ft away.

Uniformity:
By far the best OLED I’ve ever owned. My C2 has horrible vertical jailbars visible in all dark content. The G5 is close to perfect: I see a small vertical bar on the right side of the screen, but I haven’t seen it in real content yet. Haven’t noticed any DSE as well.

Tint:
There’s a slight blue/magenta tint when viewed at an extreme angle, but it’s not too bad.

[en6ads]

Each workflow creates its own temp folder so it's fine, and I do it all the time.
Just don't use two instances of the same workflow.

I've had this issue, also.

Is there a way we can bind each task (not workflow) to its own temp folder? Have a unique name for the temp folder such as UUID?

Thank you.

[Kuler087]

It should already be able to create a new temp folder when it already exists, but other commands might fail, so I never risk using two instances of the same workflow

:: make and set temp folder
if exist "%temp_folder%temp.folder_new" if exist "%temp_folder%temp.folder" rmdir /Q /S "%temp_folder%temp.folder_new" & rmdir /Q /S "%temp_folder%temp.folder"
if exist "%temp_folder%temp.folder" set E1=_new
MD "%temp_folder%temp.folder%E1%"
set TEMP=%temp_folder%temp.folder%E1%\

[en6ads]

This may sound like a dumb question, but I'll ask it anyway:

Are there any P7 MEL disks where the MEL layer contains some (small) amount of data that would prevent a perfect conversion from P7 to P8.1? Can I assume all MEL layers are redundant and can be deleted, creating a perfect conversion to 8.1?

Many thanks.

[en6ads]

Perhaps another dumb question:

Is it possible to convert a 1080p SDR AVC blu ray rip to Dolby Vision Profile 9?

Thanks.

[Kuler087]

Are there any P7 MEL disks where the MEL layer contains some (small) amount of data that would prevent a perfect conversion from P7 to P8.1?
No, MEL can always be converted to P8 without loss.

Is it possible to convert a 1080p SDR AVC blu ray rip to Dolby Vision Profile 9?
not sure if I ever heard of P9 ? maybe you meant profile 8.2?
SDR in an HDR container is just 100 nits, so it wouldn't be useful.

[domyx]

Sorry if this has been asked before, I did some research but couldn't find the answer in the forums or the YT channel. Hope I tried the correct search keywords :)

I have two remuxes of the same release:
- Remux 1: DV7 (BL + FEL)
- Remux 2: HDR10
The simple HDR10 version seems to have better base layer detail than the DV7 BL (even with FEL baked in). I’m trying to see whether I can extract and convert the RPU from the DV7 version to DV8, then mux it with the HDR10 base from remux 2.

Here's what I’ve verified so far:
- The color grade appears to be the same. I used the workflow 7-2 and compared screenshots from the DV7 BL+FEL and HDR10 => gradients and brightness look the same.
- Scene cuts are also aligned (same frame numbers)

Given that, is it safe to proceed with extracting the RPU from the DV7 remux, and muxing it onto the HDR10 base (and converting it to DV8 in the process)? Are there any other checks I should do? Am I completely misunderstanding something here?

[Kuler087]

If the grades match, then no problem.
You may want to run a complete pixel measurement in MadVR (6-2) if you want to perform a true grade check. Just a bunch of random screenshots can be misleading sometimes.

[domyx]

If the grades match, then no problem.
You may want to run a complete pixel measurement in MadVR (6-2) if you want to perform a true grade check. Just a bunch of random screenshots can be misleading sometimes.

I tried to do that, and there is actually a slight difference: both plots look the same, but the plot from the HDR10 remux is ~60nits higher on. The thing is that I think the plot of the DV7 remux is done from the BL (so without the FEL baked in), whereas the screenshots (made with 7-2) are made with the FEL baked in. At least that's my understanding, am I on the right track?

I can post the slowpics if that's easier, not sure if it was allowed here.

Note: oh, and thanks for those amazing scripts, it led me to start understanding DV and HDR a lot better!!

[Kuler087]

small difference in the plot is normal. You probably should compare the 6-2 peaks with the L1 plot in case FEL expands the brightness, eventhough you did the fel baked screenshot comparisons.

[domyx]

I believe the FEL is expanding the brightness. However, to my eyes, the overall brightness of the DV7 remux with FEL baked in (i.e., BL+FEL) looks about the same as that of the HDR10 remux. Which means the DV7 BL would be darker than the HDR10 remux (by ~60nits? which would explain the difference I see with the 6-2). Is that possible?

Another thing that makes me think the DV7 base layer is darker than the HDR10 version is when I compare both using VapourSynth/vspreview without DoViBaker (so only the base layer is shown). In that case, the HDR10 remux clearly appears brighter.

[Kuler087]

which movie? if FEL expands brightness, but your HDR10 6-2 measurement looks the same, then it probably can't be done. You may want to bake FEL to prores (8-2-3) and then measure it in 6-2.
you can post slowpics comparisons here. Just don't refer to pirated content.

[domyx]

I feel like I am missing some understanding to decide correctly. I thought that the RPU was being applied to the BL+FEL, so it would be good if the brightness of BL+FEL was the same as the brightness of the HDR10 remux (although the brightness of the BL alone is darker than the HDR10).

Here is the slowpics (https://slow.pics/c/EeC653AK) with:
- remux A (DV7 FEL) 6-2
- remux B (HDR10) 6-2
- remux A (DV7 FEL) 6-3
I am currently running 8-2-3 on remux A (DV7 FEL), and I will run 6-2 on it when it's done, but it will take some time.

Movie is "Now you see me".

[Kuler087]

yeah, that's not the same 6-2 madvr brightness at all. What I meant by small difference is ok: https://slow.pics/c/mhgoENHh

[domyx]

Ok, so I think I got it wrong with my first compare: I was comparing the DV7 remux directly.
I just did a 8-2-3 (bake FEL to prores) and a 6-2 measurement, and compared that with a 6-2 of the other HDR10 remux. The slowpics seem to indicate it's pretty close: https://slow.pics/c/U7BtfnbW
Similar I think to what you mean by "small difference".

[Kuler087]

Correct. It means the studio did not deliver a trim pass for the HDR10 base layer as they did for the FEL release.
So the P7 RPU can be converted to P8 and injected into the other HDR10 layer.

[domyx]

Thanks a lot for helping me understand that, I feel more confident in my understanding now.

[en6ads]

Hardware encoders for encoding (8) jobs.

I have read quite a few posts that state don't use hardware encoding for storage, leave it for live streaming. But I don't have any data to verify this.

Do encoding jobs such as baking FEL (8-2-1) need CPU encoding for the best quality? Is there a difference in quality between NVidia nvenc encoding and software encoding? What about Intel's encoders such as QSV on the A310 / A380? Could they be enabled for testing?

Many thanks.

[Kuler087]

The script supports only CPU or Nvidia encoding.
The quality will always be better using the CPU, but it's up to you. Do some comparisons, and if Nvidia is good enough for your needs, go for it and save tons of time.

[en6ads]

The script supports only CPU or Nvidia encoding.
The quality will always be better using the CPU, but it's up to you. Do some comparisons, and if Nvidia is good enough for your needs, go for it and save tons of time.

I'd love to test this but I don't have an NVidia GPU. I do have an Intel A380. Would it be possible to enable Intel accelerated encoding? Many thanks.

[Kuler087]

No sorry mate, I have no plan to change the encoder's workflow.

[TR-9970X]

I'd love to test this but I don't have an NVidia GPU. I do have an Intel A380. Would it be possible to enable Intel accelerated encoding? Many thanks.

I was expecting that answer from Kuler....

He'd probably need to buy an Intel GPU, and why should he, when nVidia is the way to go.

I bought a 4080 Super (with the 9950X3D) just to use DS better, I now also have a 5070 Ti, (with the 7970X) but I've yet to use it with DS...be interesting to compare.

What CPU have you got ??

[Z2697]

Changing the encoding process is easy enough, I guess the problem is that without an Intel GPU, there's no way to ensure the functionality.
But there's a potential collaborative development + testing possible.

But if goal is just "to see how it looks", I think it's not necessary to go through the script, just take some regular sample and encode them. (assuming you know how to use qsv)
Spoiler alert, it won't be as good as x265, but with enough bitrate, the difference might not be noticeable.

[dkangel]

hello everybody

can someone tell me what to do with this message in dvscript ?

L5 offset: TOP=N.A., BOTTOM=N.A., LEFT=N.A., RIGHT=N.A.

do i to correct it ?

thank you for your knowledge

[Kuler087]

When L5 is missing, the script will force you to edit it.
Use avspmod to measure your video letterbox. see: https://youtu.be/J06dUCePdd8?t=69

if your video has no black bars, enter 0 for left-right-top-bottom
If your video is cropped, you have nothing to do; the script will automatically add L5 to 0.

[dkangel]

i never got the message to force to correct it but i understand how to cerrect it :)
it's normal from a retail remux ?
in my case i have black bars (277/277)

[Kuler087]

Right, the script will force you to edit L5 only when using Workflow 1 with two inputs. For a single input, you have to select it manually.
Missing L5 on Blu-ray discs is rare. On streaming platforms like iTunes, it's missing from all titles.
L5 only matters for a few players though, specifically, CPM A14 devices or a TV's internal player so don't bother with L5 if you use other media/bluray players.

[dkangel]

ok so i will correct it because my plans are to buy ugoos player to play fel correctly

thanks for your help