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

[lemaireus]

Kuler, I tried encoding with DS today but simply have not been able to get the script firing. I am using v3.0.6, and the script is DoVi_Scripts_MKV.bat. The only change I made in the script was to change the CRF value of an x265 encode to 24. The second time, I tried doing a two-pass encode. In both the cases, I run into the same error:

https://i.postimg.cc/26H1n7Rc/Screenshot-2025-02-25-184423.png

Avisynth+ is installed correctly as are all the other tools. What might I be doing wrong?

[Kuler087]

I don't know. try installing this AVS+ https://github.com/AviSynth/AviSynthPlus/releases/download/v3.7.3/AviSynthPlus_3.7.3_20230715_vcredist.exe

https://i.ibb.co/Xr6b3xG6/Windows-Terminal-xu4-Soy-QFMB.gif

[lemaireus]

Thank you for the suggestion, Kuler, but it didn't work. In fact, I deleted the tools folder, and downloaded it afresh, but still getting exactly the same error. It's probably some sort of a hidden problem specific to my computer. Very frustrating.

[lemaireus]

This is the error I get every single time:

avs+ [FLAW]: Error loading file: Script error: Invalid arguments to function 'z_ConvertFormat'.
(D:\DoVi.Scripts\temp.folder31\script.avs, line 5)
x265 [FLAW]: unable to open input file <D:\DoVi.Scripts\temp.folder31\script.avs>

Error: The file 'D:\DoVi.Scripts\Gladiator_II_2024_4K_002_DV-FEL_to_DVP8-HDR10.hevc' could not be opened for reading: open file error.

[Kuler087]

yeah that's strange, it looks like your system doesnt like z_convert. First time seeing this error and I helped many people with this workflow.

try this version, I replaced z_convert by another function.

https://drive.google.com/file/d/1hmi7vuoz6JWd2j5mVVQA1IcqZ43yJ70d/view?usp=sharing

[lemaireus]

Thank you for the test script, Kuler. Unfortunately, this also generates an error, seemingly a different one but in fact the error involves exactly the same steps:

avs+ [INFO]: AviSynth+ 3.7.3 (r4003, 3.7, x86_64)
avs+ [FLAW]: Error loading file: Script error: Invalid arguments to function 'ConvertToYUV420'.
(D:\DoVi.Scripts\temp.folder31\script.avs, line 5)
x265 [FLAW]: unable to open input file <D:\DoVi.Scripts\temp.folder31\script.avs>

Error: The file 'D:\DoVi.Scripts\Gladiator_II_2024_4K_002_DV-FEL_to_DVP8-HDR10.hevc' could not be opened for reading: open file error.

I would believe that mine is fairly standard computer: Windows 11 machine with Core I5 Raptor Lake. Can't figure out what the source of the problem might be. It looks like an avsynth+ issue of some kind.

[lemaireus]

I usually use Ripbot for encoding and that also uses avisynth+ but that has never given me this particular error, so I can't figure out what the source of this problem might be. The error is consistently "invalid arguments to function . . . " in line 5 of the avisynth script.

[Kuler087]

no idea why it says that for you. Try to open the script (temp folder) in avspmod.

[lemaireus]

This is how the avs script looks in avspmod:

https://i.postimg.cc/FzHVkb6n/Screenshot-2025-02-26-182957.png

However, when I try to run avspmod, before loading the avs script whose screenshot I've shared above, it gives me this error message:

https://i.postimg.cc/jdChsXPD/Screenshot-2025-02-26-183011.png

[Kuler087]

the script is incomplete, so it fails at indexing either with ffms2 or dgindex

did you install python? Watch the script installation tutorial.

[Kuler087]

it looks like the script failed to detect if you have an nvidia GPU or not...

try this: https://drive.google.com/file/d/128gq8aDUTKA_aT7SQsM9dkjA1EP1sosR/view?usp=drive_link

[lemaireus]

Voila! and it worked, finally! Thank you, Kuler, for helping out. The custom script you shared is doing the job even as I type this message. I'll post an update after the encoding finishes (will take a while).

However, the problem could not have been the script failing to detect nvidia GPU. Each time I tried with the .bat files from 3.0.6, those did detect that I do not have an nvidia GPU ("No Nvidia GPU detected, enabling FFMS2 decoding"). Python is also correctly installed. Still cannot figure out the issue.

[Kuler087]

It was my fault, sorry. I introduced a typo two versions ago...
It was detecting correctly that you didn't have an Nvidia GPU but there was a typo in the variables.

[lemaireus]

Thank you, Kuler. I just got my first DS encoded rip (a ten minute test clip).

One question though: how do I get DS to keep FEL/MEL and not only the 8.1 RPU? If I am not concerned with compatibility issues and if I wish to keep DV profile 7.6 along with the FEL in, for example, Gladiator II, are there settings that I might tinker with which will allow me to do that? The reason I ask this question is because as far as converting the RPU to 8.1 and injecting it back into an encode is concerned, DDVT tools does a good job. However, as you mentioned on that forum, DDVT probably cannot bake an FEL/MEL accurately. Can DS do that?

[Kuler087]

Not possible. You cannot keep the EL if you re-encode the BL, that's the whole point.
The dovi baker process in workflow 8-2 will bake the 12bit EL stream in the encoded BL so you get a P8 + HDR10 BL that is the same quality as the original P7 fel playback on a bluray player.

[lemaireus]

Thank you for the clarification, Kuler, and, once again, thank you for helping out with the script. Interacting with you is always such a pleasurable learning experience.

[Evilmatcher]

I came across this nice tool pack (very impressive work behind it), when I discovered it was possible to convert my HDR10+ movies to DV, so I could benefit from the dynamic metadata on my LG G4. I have now converted my two HDR10+ movies (American Gangster and Carlito’s Way) and a few others HDR10 movies to DV P8 with your script. I must admit I find it difficult to see the changes. I read a post somewhere the other day (sorry I can’t remember where, but not here) about a guy who had converted his copy of Inception to DV8, where he was very impressed and the movie couldn’t look any better, so I started with Inception. I haven’t watched the complete movie, but has tried to compare the HDR10 and DV8 versions by viewing a couple of minutes of each version until around the 15 minutes mark, which I find very difficult to spot any difference. I think I see some small brightness changes here and there, but not a day and night changes.

I am using the 3-1 workflow with the MDL primaries set to P3 4000 nits (taken from MediaInfo) and using the auto level 5.

I am using a Ugoos AM6+ with CoreElec (and have also tried my X800m2) through a Denon X4800 receiver with ultra high speed hdmi 2.1 cable to receiver and to TV. The Ugoos is cable connected to a Synology DS920+ NAS.

LG G4 77” (not calibrated) but using the picture settings from https://www.flatpanelshd.com/review.php?subaction=showfull&id=1727259776#4. I have tried to tweak these settings with the Spears & Munsil UHD calibration disc earlier on, which only updated these settings marginally.

I have converted the HDR10+ movies above, Inception, Black Hawk Down and the two Blade Runner movies, but stalling on my remaining +50 HDR10 movies as it is quite a time consuming exercise (9-14 hours per movie), which I can’t “justify” for a marginally effect.

Have I been too optimistic and set my expectations too high? Any advise or further clarifications on what the tool pack should do will be very helpful.

[Kuler087]

The G4 is a bright TV, so unless the content is constantly higher than your TV's 1500nits peak brightness, there will not be much difference between HDR10 and DV in most movies because no tone mapping is required. TVs get brighter every year, so eventually, DV will be useless.
PS: You should never rely on any form of auto-L5 or auto black border detection; always do the measurement manually for proper results. This is even more important now that the ugoos AM6B+ and CPM 13 support it.

You can look at the brightness plot (https://drive.google.com/drive/u/1/folders/154fBNllwOHL4Lckc7wDV8QKFJwFxnDt-) and decide if it's worth the effort.
You can also create metadata comparisons (https://drive.google.com/drive/folders/1g5I-z_sJmVu-SAIPNiiSlcdMiy2ka0mf) with workflow 7-4 to see the tone mapping effect of DV on your TV.

EG, this is on a 400nits display (your G4 is 1500nits)

https://i.ibb.co/MkMKN6d9/Application-Frame-Host-BD943-QYk-QW.gif

[Evilmatcher]

The G4 is a bright TV, so unless the content is constantly higher than your TV's 1500nits peak brightness, there will not be much difference between HDR10 and DV in most movies because no tone mapping is required. TVs get brighter every year, so eventually, DV will be useless.



Aaaaaaaaaaaaaaaah - that make sense.
Thanks for the clarification.

[Korbus]

Hi RESET, everytime I try to use workflow 7-4 or workflows of Mode 6 and the input is a DV mkv (profile 7 or 8), i get an error such as in the picture attached (basically the ffmpeg windows displays all kinds of weird symbols and crashes).

[Kuler087]

What the hell lol. No 2nd window is supposed to pop.
Can you show me the log (https://justpaste.it/) with ''@echo on'' (line 2)

[Korbus]

Here it is: log (https://justpaste.it/gjtsq)

[Kuler087]

strange, it seems ffmpeg fails to pipe/extract the RPU. Did you change your tool folder ffmpeg.exe version? Try replacing it, maybe it got corrupted when you extracted the tools pack?
You could try to run the cmd that fails manually to see if it's the same.

"D:\Downloads\DoVi.Scripts\tools\ffmpeg.exe" -i "D:\Downloads\DoVi.Scripts\temp.folder216\Blade Runner (1982) 4K DV.mkv" -c:v copy -bsf:v hevc_mp4toannexb -f hevc -loglevel error -stats - | "D:\Downloads\DoVi.Scripts\tools\dovi_tool.exe" extract-rpu -o "D:\Downloads\DoVi.Scripts\temp.folder216\RPU.bin" -

[Korbus]

I have replaced ffmpeg.exe (I had not change that file, anyway), same thing, the second window popped. If I run the command manually (I suppose by opening the Command Processor), but only if I did not delete the temp folder after the ffmpeg.error, it runs very quickly and says "Metadata reordered", as in the screenshot attached, but apparently nothing else happens (in the temp folder, however, the RPU.bin that was showing 0 bytes now appears as ~5,7 MB).
If I run the workflow and close the second window quickly enough (have to end the process), it seems to go further, but still ends in error (see log (https://justpaste.it/eu949)).
L.E. Have to mention that other workflows using ffmpeg.exe work normally, otherwise my 8-2-1 or 3-1 workflows would end in error, but they do not.

[Kuler087]

so the cmd works manually but not when you run the script... I have no idea why, and I can't reproduce it, so there's not much I can do about it sorry.
Maybe an Admin right issue? your manual CMD is with admin rights, the script is not otherwise drag and drop doesn't work. Did you try the manual / full mode ?

I tried with the same filename/path and the same output path as you, and it works fine on my end so I don't know:

https://i.ibb.co/Hf1hnNjf/Windows-Terminal-f-Sd-K9-Qzq-Uj.gif

[Korbus]

I have just checked and ffmpeg.exe was set to run in Administrator mode. Once I have unchecked that, the flow proceeded just like in your .gif, and now seems to be running the encode. However, before starting the encode, it threw this error, for each of the 10 frames:
"Measure Frame: 2000 (1 of 10)
Traceback (most recent call last):
File "D:\Downloads\DoVi.Scripts\tools\Measure_HDR.py", line 1, in <module>
import cv2
ModuleNotFoundError: No module named 'cv2'"

Let's see if it finishes the encode...
L.E. It did finish the encode, resulting in an .mp4 file. Thanks a lot for your help, I'll now check other workflows that failed before due to ffmpeg.exe (Mode 6) and report back.

[Kuler087]

That's a missing Python library needed to measure the frame maxcll/fall.

Open a cmd window and type:
pip install opencv-python


other python library required for other specific workflow:
colour-science, scikit-image, matplotlib, numpy, colour, headless, PyQt5

[Korbus]

Yes, I have just played the .mp4 and, on my TV (Sony A95L), it showed zero difference between no metadata, L1+L3 or L1+L2+L3+L8.
I have installed the libraries you've listed and started again WF 7-4. Now encoding, no error messages this time. Anxious to see the result :)

[Kuler087]

the opencv step only matters for the frame actual maxcll/fall measurements for the OSD.
It doesnt matter for the DV metadata comparison. If you don't see any difference between no metadata/with metadata, it's because the random frames selected are all within your TV's capabilities.


this is the metadata you get when you select the auto-mode for this movie. I think the A95 has a 1500nits target.
https://i.ibb.co/GQjQ3YCh/Blade-Runner-1982-The-Final-Cut-HDRg-DV-Metadata-Levels-Comparison-CMv4-0-Do-Vi-L1-PLOT.png

[SeeMoreDigital]

I have replaced ffmpeg.exe (I had not change that file, anyway), same thing, the second window popped. If I run the command manually (I suppose by opening the Command Processor), but only if I did not delete the temp folder after the ffmpeg.error, it runs very quickly and says "Metadata reordered", as in the screenshot attached, but apparently nothing else happens (in the temp folder, however, the RPU.bin that was showing 0 bytes now appears as ~5,7 MB).
If I run the workflow and close the second window quickly enough (have to end the process), it seems to go further, but still ends in error (see log (https://justpaste.it/eu949)).
L.E. Have to mention that other workflows using ffmpeg.exe work normally, otherwise my 8-2-1 or 3-1 workflows would end in error, but they do not.

Are you guys trying to add 'fake' Dolby Vision to Blade Runner?

[Korbus]

@RESET

Oh, OK, that makes sense. Thank you!

And one more unrelated question, please. When I play this test file (FEL TEST ST DL P7 CMV4.0 4000nits V3) on my Ugoos AM6B+ (running cpm build, but CE vanilla does the same), it never progresses past L2 (CMv2.9) stage. No L1 or L8 (CMv4), the video keeps playing, but the image stays unchanged. I have always wondered why and debated with a colleague with a LG B9 TV, who said that on his TV internal player, that file plays correctly.

@SeeMoreDigital

Yes, I 've just played, out of curiosity, with the different workflows from DS and in some cases really liked the result, in others there was no noticeable difference, in others it was worse than HDR10.

[Kuler087]

Are you guys trying to add 'fake' Dolby Vision to Blade Runner?

Fake? DV metadata can be generated freely using the official Dolby algo for any HDR movie.

https://customer.dolby.com/content-creation-and-delivery/dolby-vision-professional-tools-v550

[Kuler087]

And one more unrelated question, please. When I play this test file (FEL TEST ST DL P7 CMV4.0 4000nits V3) on my Ugoos AM6B+ (running cpm build, but CE vanilla does the same), it never progresses past L2 (CMv2.9) stage. No L1 or L8 (CMv4), the video keeps playing, but the image stays unchanged. I have always wondered why and debated with a colleague with a LG B9 TV, who said that on his TV internal player, that file plays correctly.

Yes it's normal because the EL is homemade, and it crashes after the L2 metadata.
However, this test file's purpose is to check FEL support. There are better test files (https://drive.google.com/drive/u/1/folders/1nMz95KqgkO95EDGrYhrk-Ifx6vRYDEfs) for testing cmv4.0 or any RPU metadata.
B9 doesn't crash because it doesn't decode the EL, I guess?

in others it was worse than HDR10.
Dolby Level 1 metadata is a non destructive process. It cannot be worse than HDR10 as long as your TV has a proper target brightness of course.

[Korbus]

Thanks for the explanation!
By "worse" I've meant that, in some movies, DV tended to crush the blacks, compared to the HDR10 (The Equalizer, for example). Of course it did not destroy anything, as the RPU can be easily removed or the DV can be turned off on the player and the movie can be played as the creators intended.

[SeeMoreDigital]

@SeeMoreDigital

Yes, I 've just played, out of curiosity, with the different workflows from DS and in some cases really liked the result, in others there was no noticeable difference, in others it was worse than HDR10.Is the resulting encoded considered to be MEL of FEL?

[Korbus]

Neither. It's a 8.1 profile (no EL, only BL+RPU), so it's the equivalent of MEL.

[Kuler087]

Thanks for the explanation!
By "worse" I've meant that, in some movies, DV tended to crush the blacks, compared to the HDR10 (The Equalizer, for example). Of course it did not destroy anything, as the RPU can be easily removed or the DV can be turned off on the player and the movie can be played as the creators intended.

Do you say that based on a DV metadata comparison from 7-4?

if yes, here is some explanation:

No metadata does not equal HDR10. No metadata just disable the TV tone mapping and hard clipping will be visible if the content exceeds your TV capabilities. HDR10 will also dim/tone map the image for that same shot according to its static tone mapping curve.

Also, don't forget that DV is shot by shot, and workflow 7-4 selects only one still image of that whole shot.

[Kuler087]

Is the resulting encoded considered to be MEL of FEL?

as Korbus said, P8 DV, but what he was trying to make work is not the workflow that can generate DV from HDR10.

We are talking about workflow 7-4 that can create DV video comparisons of ''original'' or ''generated'' DV rip. It can select, automatically or manually 10 different frames, and the script will encode those images in a loop and altern between the Dolby Vision metadata on and off to see the Level 1 and the trims effect on the image as shown in the gif posted here (https://forum.doom9.org/showpost.php?p=2015591&postcount=1018)

[Korbus]

No, just by watching a particular scene in that movie (The Equalizer). Around the 00:18:53 timestamp Denzel starts to walk with the girl across a bridge (it's a night scene). That scene (1-2 minutes) it's a tad darker and exhibits a light black crush in my DV-cooked remux (WF 3-1, CM algo, tried both with target MDL 1000 and 4000 nits). Perceptually, it's darker on the DV version, although in daylight or artifficial light scenes looks great.

[SeeMoreDigital]

Neither. It's a 8.1 profile (no EL, only BL+RPU), so it's the equivalent of MEL.I see...

I created a 4K UHD Blu-ray MEL encode of Blade Runner myself a few days ago using TSmuxerGUI and by adding the (green) secondary MEL stream originally from James Bond Casino Royale. It plays fine in my OPPO UDP-203. Overall it looks brighter and in some areas a bit grainier.

[Korbus]

Oh... that doesn't sound right to me, but I guess I'll let RESET chime in...

[Kuler087]

I see...

I created a 4K UHD Blu-ray MEL encode of Blade Runner myself a few days ago using TSmuxerGUI and by adding the (green) secondary MEL stream originally from James Bond Casino Royale. It plays fine in my OPPO UDP-203. Overall it looks brighter and in some areas a bit grainier.

wow, don't do that. You cannot take the metadata from one movie and mux them to another movie , that's not how DV works. Casino royal is a 500nits grade, and Blade Runner 1982 is much brighter (1000nits +), so your experiment is telling your TV to tonemap a 1000nits+ movie like it was a 500nits movie. So less tone mapping = brighter image of course but also more clipped.

You can generate genuine DV metadata with dovi_scripts workflow 3-1 (https://www.youtube.com/watch?v=jBqbG5XM54g) or with Resolve(faster) (https://www.youtube.com/watch?v=FVSh3oGqfXY). Your oppo supports Profile 8 TS/M2TS rip just fine

[SeeMoreDigital]

wow, don't do that. You cannot take the metadata from one movie and mux them to another movie , that's not how DV works. Casino royal is a 500nits grade, and Blade Runner 1982 is much brighter (1000nits +), so your experiment is telling your TV to tonemap a 1000nits+ movie like it was a 500nits movie. So less tone mapping = brighter image of course but also more clipped.

You can generate genuine DV metadata with dovi_scripts workflow 3-1 (https://www.youtube.com/watch?v=jBqbG5XM54g) or with Resolve(faster) (https://www.youtube.com/watch?v=FVSh3oGqfXY). Your oppo supports Profile 8 TS/M2TS rip just fineDon't panic, it was just an random experiment to see if it was possible to create a blu-ray disc compliant mux complete with a secondary video stream...

The entire secondary MEL stream from (the 138 minute) James Bond Casino Royale movie was less than 150MB. By contrast the secondary MEL stream from 2001：A Space Odyssey was over 2GB.

EDIT: Many thanks for confirming that my OPPO can play 'Profile 8 TS/M2TS', I didn't know this :)

[foxman679]

The G4 is a bright TV, so unless the content is constantly higher than your TV's 1500nits peak brightness, there will not be much difference between HDR10 and DV in most movies because no tone mapping is required. TVs get brighter every year, so eventually, DV will be useless.


you wrote "so eventually, DV will be useless". Are you meaning Full DV FEL with 12bit Diff-Stream or only MEL-DV?

[Kuler087]

Yeah I was referring to the brightness/gamut tone mapping.
I guess FEL will always be useful to achieve 12-bit or fix stupid studios that can't compress the HDR10 BL properly.

[Kontrarian]

@Kuler087, thank you for everything you do, your work is awesome!

I have a question regarding the Ant-Man (2015) RPU from your GD.

This is the RPU L2 plot from your P5 to P8 RPU bin file:
https://i.ibb.co/fVgdCxwq/Ant-Man-2015-P5-to-P8-Do-Vi-L2-PLOT-100nits-trim.png

It's identical to what I pulled from another source (except for the Shots):
https://i.ibb.co/dwCtd4KX/Ant-Man-2015-UHD-Blu-Ray-2160p-True-HD-Atmos-7-1-DV-HEVC-HYBRID-REMUX-Fra-Me-STo-R-RPU-Do-Vi-L2-PLOT.png

The ms.weight looks crazy on both of them.

So I grabbed a different WEB DV and plotted that one and got this:
https://i.ibb.co/0jCcMSQx/rvkd-ant-man-2015-dv-2160p-web-h265-RPU-Do-Vi-L2-PLOT-100nits-trim.png

I'd appreciate your thoughts on the differences?
Thank you.

[SeeMoreDigital]

Hi @Kuler087,

Maybe you're able to answer a question which came up on the Blu-ray.com forum a while ago.

With regard to the MEL encoded secondary streams, the bit-rate/file size on some discs significantly larger than on others.

For example, the MEL encoded streams for disc such as: 'James Bond: Casino Royale' (138 mins), 'Birds Of Prey' (109 mins) and 'Dune [1984]' (136 mins) have bit-rates at around 125kbps (around 150MB). But the MEL encoded streams on discs such as: '2001：A Space Odyssey' (143 mins) and Gemini Man (117 mins) are huge at around 2150kbps and 2750kbps respectively (around 2.35GB)!

Is there any particular reason why there's such a huge difference? Indeed, they're more than 16 times larger?


Cheers

[Kuler087]

I have no idea why some studios decide to use the larger MEL bitrate, but it doesn't matter. The quality is the same whether the bitrate is 125kbps or 2mbps; it's just a waste of space, IMO. I always convert MEL to P8 since it has better support on devices.

[SeeMoreDigital]

I have no idea why some studios decide to use the larger MEL bitrate, but it doesn't matter. The quality is the same whether the bitrate is 125kbps or 2mbps; it's just a waste of space, IMO. I always convert MEL to P8 since it has better support on devices.It is weird that's for sure...

So, is it possible to create a Blu-ray spec compatible MEL secondary video stream using your scripts, that can be muxed together with the primary video stream?

Has anyone created one for say, Blade Runner (1982) or Blade Runner 2049 that I can use to test with my OPPO UDP-203?

[Kuler087]

Workflow 4-1 (or tsmuxer) will mux any raw hevc you feed it, but the output will be incorrect when the player will decode the layers because of different NAL and IPB frames order (the same reason why you can't re-encode the BL and mux EL). Similar to all these misauthored discs (https://docs.google.com/spreadsheets/d/15i0a84uiBtWiHZ5CXZZ7wygLFXwYOd84/edit?gid=671480574#gid=671480574) that can't be muxed/played in P7 properly.

What you want is profile 8 DV. Here's a P8 generated RPU (https://drive.google.com/file/d/1oW2KaCNv5HrFBQnQ7aJdQyStrp5nYNNV/view?usp=sharing) for blade runner if you want to experiment.