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

[TDS]

Originally Posted by DMD

You can try ProRes 422HQ or 4444 XQ using the free Voukoder plugin.
https://www.voukoder.org/forum/threa...oogle_vignette
https://github.com/Vouk/voukoder

Thanks again, DMD.

Got the latest build & then I had to get the connector, so I will see what that does for me.

[DMD]

Originally Posted by TDS

Thanks again, DMD.

Got the latest build & then I had to get the connector, so I will see what that does for me.

Installation instructions:
To use the classic version of Voukoder, two components must be installed:

The Voukoder application and
The connector plug-in that connects Voukoder to your favorite NLE.

1. Install the main Voukoder application.

2. Extract the contents of the file "resolve-connector-x.xx.x.zip" to:
C:\ProgramData\Blackmagic Design\DaVinci Resolve\Support\IOPlugins

3. If the "IOPlugin" directory does not exist, create it.

4. Done


Settings:




Selection options ProRes 4222 HQ, 4444 HQ and 4444 XQ







In the audio section, select the Uncompressed PCM codec, as if different, StaxRip can block in the Demux audio section.




Selection container mov.

[TDS]

Originally Posted by DMD

Installation instructions:
To use the classic version of Voukoder, two components must be installed:

The Voukoder application and
The connector plug-in that connects Voukoder to your favorite NLE.

1. Install the main Voukoder application.

2. Extract the contents of the file "resolve-connector-x.xx.x.zip" to:
C:\ProgramData\Blackmagic Design\DaVinci Resolve\Support\IOPlugins

3. If the "IOPlugin" directory does not exist, create it.

4. Done


Settings:

Yes, I done most of that, a few times

I'm curious that a YUV (10bit 4:2:0) isn't displayed, there's only YUV (High Bit Depth)
I've delivered one at both settings and the 8 bit is a larger file size that the high bit..

A lot more trial & error, and thanks again for your help

[wyup]

Originally Posted by Kuler087

That is not how it works. As soon as you change something in the video, that's re-grading regardless of how bright your source is(100, 200, 500nits).
There is no difference in the workflow and any change you make must be monitored shot by shot otherwise you're raising everything, even the non-highlight pixels (EG: white objects or source clipping that will look off).

it does not. You're confusing clipping and maximum brightness/details your TV can display.

For example, this image is 3000nits &+ and appears brighter because everything is clipped by the TV but in reality, it does not respect the creator's intent and it is NOT brighter than the 800nits(my TV peak) mapped image.
Your TV cannot get brighter than its capabilities so both images peak at 800nits but the 3000nits one is clipped because my TV can't display the original brightness.

I'm not confusing with clipping because that 10,000 demo didn't clip at all on my Samsung tv. My tv reads MaxDML HDR10 metadata and tonemaps it to its capacity. It has two curves: one for 1,000 to 4,000 nits, and another for 4,000 to 10,000 nits. 0 to 1,000 is mapped directly as is.

Can I obtain a 400, 600 or 1,000 nit tonemap from the 10k nit original with your scripts or from your drive? This way I can compare between them and check. I'm not sure, but I have the impression that low nit videos, such as 400 to 600 nits watch darker than a 4k or 6k nit. Yes, i know they should watch similarly, but I'd like to check up.

However, a 400 to 1,000 up tonemapping would make sense because it would benefit from the full capacity of my tv. Otherwise, 600 nits would be wasted IMHO.

Fair enough, bringing up brightness with Resolve might change the grade already, but there must be a way of doing it 'transparently', or 'non destructively'. I.e: when I use brightness or peak brightness control on my tv, the image doesn't change much per se, it remains largely the same. It is a similar process to watching SDR: it is mapped to the tv capacity and blacks remain anchored. I think tonemapping algorithms are for down-tonemapping mostly, because you must make concessions and priorities. For up-tonemapping, it is a matter of scaling up brightness, nothing is lost.

By the way, I wonder how they grade movies at 10,000 nits: they must use sunglasses because for one, mastering monitors of that capacity must be extremely expensive, and who can watch luminance that high? I presume these are up-tonemapped for mastering.

For example: have a look at this 3-minute 200 Mb (no Dolby Vision, just HDR10) clip extracted from a download of a 2023 30th Anniv. HDR10 remaster of The Fugitive (1993) film. It has a maxCLL of 5,800 nits and average of 1,500 nits. Since it comes from celulloid, it is impossible the negative had luminance this high: it must have been up-tonemapped. The whole movie has a maxCLL of 6,800 nits, so a mastering monitor of at least that much must have been used, if not much less and then up-mapped.

plot
clip

[Kuler087]

Can I obtain a 400, 600 or 1,000 nit tonemap from the 10k nit original with your scripts or from your drive?

yes, I've added it in the beta version the day after you asked but I haven't tested it yet. The SM disc already have a 1000/600 nits version of the montage btw.

line 169:
::Choose the trim target to deliver in 8-2-5. Choicies are: 100nits=1 / 600nits=28 / 1000nits=49 (default = 1 )
set CMOFFLINE.TARGET=1

Dolby CM_offline has the same prores slow performance bug as CM_analyze so it is much faster to do it in resolve and you have more target options.

[TDS]

Sorry for my forgetfulness, but when using 8-1-3 either DD+ or THD, I am getting a 7.1 from the original 5.1 ac3.

Is there a setting to NOT convert to 7.1 ??? I was looking around the other day, and couldn't find any relevant settings, or is it a DEE thing ??

[Kuler087]

the dee workflow is just for 7.1 audio. No point using it for 5.1, use 8-1-1

Originally Posted by Kuler087

the dee workflow is just for 7.1 audio. No point using it for 5.1, use 8-1-1

Good to know, tho

[halls]

Hi!

@Kuler087
Could you make a video on how to resync the HDR10+ metadatas and inject to normal HDR10 base layer?

As in "web streaming to to HDR10 BD rip P8 Hybrid" video.

[Kuler087]

You do that the same way as when you inject/resync DV from WEB with workflow (1-1).
https://www.youtube.com/watch?v=hVWZpat34oc

[halls]

Originally Posted by Kuler087

You do that the same way as when you inject/resync DV from WEB with workflow (1-1).
https://www.youtube.com/watch?v=hVWZpat34oc

[jayper]

If I have:
- an HDR10-only disc
- a retail P5-RPU (cmv2.9)
- a CM_Analyze generated RPU (cmv4.0)

Is there any value in running workflow 1-1 with the HDR10 base layer, and the retail P5-RPU (cmv2.9).

Then running the resulting file (HDR10 base with retail P5-RPU) through 1-1 again, this time with the CM_Analyze generated RPU (cmv4.0) to get the colorists trims from the retail P5 plus the cmv4.0 metadata from the CM_Analyze?

[Kuler087]

Then running the resulting file (HDR10 base with retail P5-RPU) through 1-1 again, this time with the CM_Analyze generated RPU (cmv4.0) to get the colorists trims from the retail P5 plus the cmv4.0 metadata from the CM_Analyze?

That would remove the retail cmv2.9 rpu and would replace it by the new cmv4.0 rpu. In other words, 1-1 always overwrite any existing RPU from the BL(input one).

The workflow to transfer metadata levels from one rpu to another is 2-6 (and now 2-3 in upcoming 2.0.9) but you cannot transfer levels from RPUs that were created from difference analysis. The Dolby 2.9 and 4.0 calculate the metadata quite differently. See this comparison
You can only transfer levels when both RPU comes 100% from the same DV analysis.

[jayper]

Originally Posted by Kuler087

That would remove the retail cmv2.9 rpu and would replace it by the new cmv4.0 rpu. In other words, 1-1 always overwrite any existing RPU from the BL(input one).

Ah, I see. I wasn't sure if that process replaced or transferred the levels. Good to know. Thanks.

Originally Posted by Kuler087

The workflow to transfer metadata levels from one rpu to another is 2-6 (and now 2-3 in upcoming 2.0.9) but you cannot transfer levels from RPUs that were created from difference analysis. The Dolby 2.9 and 4.0 calculate the metadata quite differently. See this comparison
You can only transfer levels when both RPU comes 100% from the same DV analysis.

I see, once again, I had no idea. I am glad I asked. Thanks very much.

I realize that the answer to this question probably depends on the type of TV you have, but is there a general rule as to what is better? cmv2.9 with custom trims, or generated cmv4.0 with no trims?

[Kuler087]

The only times you could replace a retail RPU by a cmv4.0 generated rpu are:

1- cmv2.9 rpu without any trims
2- a rpu with static L1/L2. (static L1 + dynamic L2 is fine)
3- if the hybrid sources are not the same grade
4- On some rare occasions, when I think the retail rpu doesn't make sense like in ''Risky Business 1983''
5- If you don't have a FEL capable device and fel expand the brightness, I guess its fine to generate DV for the HDR10 BL brightness.

[TDS]

Hi Kuler,

I have been using the latest 2.0.9 beta, and using Workflow 3 to create a "Dovi" of PotC clip.

I used 3-1 7 (4000/P3), and 3-1 20 (1000/P3) & 3-2 (no other options), and using both the new & old mp4 muxers.

And I cannot tell any difference between any of them, and both the muxers work with my new LG C4.

Not sure if there will be any difference in encoding time for the full length movie (processing atm), between 3-1 & 3-2, but I had one doing 3-1 a couple of days ago, it was going to take a L-O-N-G, the CUDA encoding countdown started at about 18 hours

Would there be any difference in encoding time, if it was just the video track (add the audio later)

Might have to start using Resolve.

Anyway, still learning, and you keep improving your workflows...

Cheers

[Kuler087]

3-2 use madVR for generating the metadata and it's not as good as the DV algo.
You could possibly get some brightness fluctuation between shots and the average_PQ metadata is much higher with madvr which will result in a darker tone-mapped image.

Here's the actual difference between metadata generated with Dolby algo, HDR10plus and Madvr: https://drive.google.com/file/d/1Pii...usp=drive_link

[TDS]

Originally Posted by Kuler087

3-2 use madVR for generating the metadata and it's not as good as the DV algo.
You could possibly get some brightness fluctuation between shots and the average_PQ metadata is much higher with madvr which will result in a darker tone-mapped image.

Here's the actual difference between metadata generated with Dolby algo, HDR10plus and Madvr: https://drive.google.com/file/d/1Pii...usp=drive_link

So does that explain why 3-2 is a faster process that 3-1 ??

Can this be done with Resolve ??

[Kuler087]

Of course, 3-2 will always be faster than 3-1 and faster than Resolve as well because it can create the metadata straight from the HEVC stream with madMeasureHDR.exe or HDR10plus.

In 3-2:
if present, will convert HDR10plus to DV
if no hdr10plus, it will use madvr to measure the video and convert the madvr measurement file to DV.
in both cases, no prores transcoding and no slow performance (cm_analyze.exe) so that's why it is so fast but less accurate than using the official Dolby Algo (the same algo the studios use).

[TDS]

Originally Posted by Kuler087

Of course, 3-2 will always be faster than 3-1 and faster than Resolve as well because it can create the metadata straight from the HEVC stream with madMeasureHDR.exe or HDR10plus.

In 3-2:
if present, will convert HDR10plus to DV
if no hdr10plus, it will use madvr to measure the video and convert the madvr measurement file to DV.
in both cases, no prores transcoding and no slow performance (cm_analyze.exe) so that's why it is so fast but less accurate than using the official Dolby Algo (the same algo the studios use).

OK, thanks for that, I'll have to do a couple of my own comparisons, to determine the time taken, and the "quality".

My old eye's probably won't be able to tell too much difference.

So just for my information, I am currently watching Pirates Of The Caribbean Curse of the Black Pearl, and there is a LOT of different "colour grading" through out, some scenes are quite "natural" looking, and other's are dark and "reddish", is this the grading process that you speak of that takes hours in Resolve ???