Dolby Vision now possible through MP4 Mux.

[dapope]

Just for fun, I thought I would try generating DV metadata for the Hitchcock film "Rear Window" using the new 3-1 frame by frame option. This movie was only released in HDR, with no DV. The film is in 1:66:1 (5:3) aspect ratio, so am I correct in entering the L5 offsets as:
Left L5 offset 120, Right L5 offset 120, Top L5 offset 0, Bottom L5 offset 0
?
Thank you so much!

you have to measure the black bars manually in Avspmod. I didn't generate the RPU I have but it's set to:

Code: Select all

          "Level5": {
            "active_area_left_offset": 124,
            "active_area_right_offset": 125,
            "active_area_top_offset": 0,
            "active_area_bottom_offset": 0
          }

Thank you. I got about the same result when I measured it except I had left as 125 and right as 124.

Unfortunately the file only played fine up until about the 1h20m mark. The audio kept playing but the picture froze. This is on the LG CX internal player after muxing to a .ts file. Seems like it has trouble seeking through the file, like it thinks the video is shorter than it actually is. Of course on VLC it plays fine.

[DiscoD]

@RESET_9999 On the Dolby vision excel file, you posted the L5 IMAX edits for Guardians of the Galaxy. How and where to use it? Thanks in advance.

you edit all the frames to 0,0,276,276 with a json file and then you edit all the Imax frames to 0,0,68,68 with another json file.

dovi_tool.exe editor -i D:\input.bin -j D:\276.edits.json --rpu-out D:\276.edits.bin
dovi_tool.exe editor -i D:\276.edits.bin -j D:\68.edits.json --rpu-out D:\final.RPU.bin

276/276 json

Code: Select all

{
    "mode": 0,
    "active_area": {
        "presets": [
            {
                "id": 1,
                "left": 0,
                "right": 0,
                "top": 276,
                "bottom": 276
            }
        ],
        "edits": {
            "all": 1
        }
    }
}

68/68 json

Code: Select all

{
    "mode": 0,
    "active_area": {
        "presets": [
            {
                "id": 1,
                "left": 0,
                "right": 0,
                "top": 68,
                "bottom": 68
            }
        ],
        "edits": {
            "2313-8290": 1,
            "12867-20599": 1,
            "29662-30346": 1,
            "44226-50210": 1,
            "62482-71703": 1,
            "90855-94414": 1,
            "114450-114880": 1,
            "120568-130909": 1,
            "150232-163186": 1,
            "180560-180982": 1,
            "184049-187882": 1,
            "197234-205633": 1
        }
    }
}

So if im encoding this and cropping away the 68px for the imax scenes should my RPU be 0,0,0,0 for the imax scenes and 208 for the normal scenes to make up for the difference?

[bbeny123]

I've improved mp4muxer (official dlb_mp4base) a bit.

The main improvement is a significant boost in the performance of muxing hevc files.
In the official version, after reading the input files, the mp4muxer looks like it goes idle (low CPU usage, minimal disk usage).
During the "inactivity" the program does n * n loops (where n = number of video frames) - so the longer (not bigger) the input video, the longer the idleness period. After my fixes, such a loop is made only once (by using a map-like data structure).
This fix significantly reduces the time needed to mux hevc files, the "idle" time has been reduced to almost 0. Muxing performance now depends almost only on the speed of the used disk.

Other changes:

• Fixed codec id for EL layer (codec id for the EL layer is again 'dvhe' and not 'hev1') - in this aspect, the behavior is the same as in the old 'mp4muxer_64bits' version (1.0.0)
• Support for .eac3 audio input
• Muxing progress logging

Source code, binaries, full changelog: https://github.com/bbeny123/mp4muxer

[RESET_9999]

So if im encoding this and cropping away the 68px for the imax scenes should my RPU be 0,0,0,0 for the imax scenes and 208 for the normal scenes to make up for the difference?

I don't know I never tried with cropped content but 208 for the non-imax shots sounds right. Test it on an LG TV internal player.

[dapope]

Just for fun, I thought I would try generating DV metadata for the Hitchcock film "Rear Window" using the new 3-1 frame by frame option. This movie was only released in HDR, with no DV. The film is in 1:66:1 (5:3) aspect ratio, so am I correct in entering the L5 offsets as:
Left L5 offset 120, Right L5 offset 120, Top L5 offset 0, Bottom L5 offset 0
?
Thank you so much!

you have to measure the black bars manually in Avspmod. I didn't generate the RPU I have but it's set to:

Code: Select all

          "Level5": {
            "active_area_left_offset": 124,
            "active_area_right_offset": 125,
            "active_area_top_offset": 0,
            "active_area_bottom_offset": 0
          }

Thank you. I got about the same result when I measured it except I had left as 125 and right as 124.

Unfortunately the file only played fine up until about the 1h20m mark. The audio kept playing but the picture froze. This is on the LG CX internal player after muxing to a .ts file. Seems like it has trouble seeking through the file, like it thinks the video is shorter than it actually is. Of course on VLC it plays fine.

@RESET_9999 Do you have any idea why this might be the case? Any help would be appreciated.
Also I would check if there isn't a bug in your program switching the left and right offset. I'm sure I entered the left offset as 125 and the right offset as 124, but now the metadata is saying it's reversed (like the one you posted).

[legosoff]

Hi, RESET_9999! mp4box subtitles iPhone и iPad it will work tx3g and sbtl. @jamiew just saved my life!
test.srt:lang=eng:layout=0x60x0x-1:group=2:hdlr=sbtl:tx3g

[RESET_9999]

@RESET_9999 Do you have any idea why this might be the case? Any help would be appreciated.

I don't know...
Network issue? HDD issue? Maybe the player choked at some point because of frame-by-frame?

Also I would check if there isn't a bug in your program switching the left and right offset. I'm sure I entered the left offset as 125 and the right offset as 124, but now the metadata is saying it's reversed (like the one you posted).

Because when you set uneven L5 offsets in resolve or cm_analyze, the dovi_tool(XML to rpu) has no way to calculate exactly which position the offsets go because the XML file works with aspect ratio values only.
This is exactly the reason why the script (1-1) has an option to edit L5 with XML inputs.

[RESET_9999]

I've improved mp4muxer (official dlb_mp4base) a bit.

The main improvement is a significant boost in the performance of muxing hevc files.
In the official version, after reading the input files, the mp4muxer looks like it goes idle (low CPU usage, minimal disk usage).
During the "inactivity" the program does n * n loops (where n = number of video frames) - so the longer (not bigger) the input video, the longer the idleness period. After my fixes, such a loop is made only once (by using a map-like data structure).
This fix significantly reduces the time needed to mux hevc files, the "idle" time has been reduced to almost 0. Muxing performance now depends almost only on the speed of the used disk.

Other changes:

• Fixed codec id for EL layer (codec id for the EL layer is again 'dvhe' and not 'hev1') - in this aspect, the behavior is the same as in the old 'mp4muxer_64bits' version (1.0.0)
• Support for .eac3 audio input
• Muxing progress logging

Source code, binaries, full changelog: https://github.com/bbeny123/mp4muxer

thanks, I'll try it.
Have you tested if the file works on older LG TVs ? (2018 or older)

[FonduemangVI]

@RESET_9999 Hey would you be able to explain what the FraMeSToR illegal L5 change at frame 203356 for GotG Vol 3 is?/why it's illegal?

[RESET_9999]

@RESET_9999 Hey would you be able to explain what the FraMeSToR illegal L5 change at frame 203356 for GotG Vol 3 is?/why it's illegal?

L5 can only change at the rpu scene cuts(refresh flag) and there is no scene cut at frame 203356.
You can export the rpu scene cut frames with my script (2-4)

[bbeny123]

Have you tested if the file works on older LG TVs ? (2018 or older)

Nope, I've tested LG C9 only.

[RESET_9999]

interesting new options in the new dolby vision tools pack:

https://slow.pics/c/5irgGGKG

L1 Analysis Tuning:
Index Title
0 Legacy CM4 L1
1 Most Highlight Detail/Most Mapping
2 More Highlight Detail/More Mapping
3 Balanced
4 Less Highlight Detail/Less Mapping
5 Least Highlight Detail/Least Mapping
Analysis tuning provides user control over the strength of filtering used during analysis. Six options are
provided that match the options found in partner tools after updating to the 2023 New CM4 L1. As the
index increases, so does the amount of filtering, lowering the L1 max.
Stronger filters may result in clipping of fine details in the highlights while reducing the amount of
mapping. By default, cm_analyze uses the 1 – Most Highlight Detail / Most Mapping
with the intent of preserving as much highlight detail as possible, with the exception of when the --
live-infinite-filter option is invoked. In this case, the analysis defaults to 0 – Legacy CM4
L1 to match the Dolby Live Distribution Processor product

Dolby is please to announce the immediate availability of the Dolby Vision Professional Tools v5.4, available now at customer.dolby.com.

New Feature Summary in this release
New L1 analysis features in cm_analyze
Added 2023 New L1 frame statistics algorithm with user-tuneable spatial filtering, new hybrid mid measurement, unified gamut analysis, and improved black measurement.
New statistics are enabled by default, and tuning can be adjusted using --analysis-tuning <arg>
See the user manual for more information
Older frame statistics are available using --legacy-cm4-l1
Improved file-based long play analysis algorithm
New temporal smoothing algorithm
New backwards compatibility metadata for frame-by-frame analysis
New difference detection to group scenes together
Added --live-infinite-filter to match analysis done in Dolby's Live Distribution Processor
Added new stereoscopic analysis for shot based and file based / frame-by-frame analysis. Right eye view can be added using --input2 <rightEye>
Added new global looks that can be added to your analysis, using --global-trims <globalLook.xml>.
See the user manual for more information
--source-format now requires both the transfer function (pq) and color gamut (p3d65 or bt2020)
• Added new functions in metafier for handling per-shot L5:
--validate performs new checks & reports
looks for and validates per-shot L5. An error will be reported if the per-shot canvas does not match the global canvas
Examines CM 4 XMLs (v4.0.2 & v5.1.0) for "illegal" legacy L1 values.
Separate count for default Target ID 1 L8 trims
--aspect-ratios <arg1> <arg2> now updates canvas of per-shot L5 and removes
See the user manual for more information
--preserve-per-shot-active preserve the current global active aspect ratio as per-shot L5 when updating aspect ratios
--global-trims <globalLook.xml> can inject a global look by applying specified trims to all shots & frames within an existing XML
See the user manual for more information
Cm_offline improvements:
Added new 300 nits P3D65 & BT.2020 targets to cm_offline
Added flag --skip-qt-st2086-min-check to permit cm_offline to continue when a QuickTime MOV's ST.2086 minimum (mdcv) does not match the Dolby Vision master display minimum
Bug fixes in this release
Fixed minor analysis bugs from 5.4-beta-1
Fixed an issue with rejecting 4:2:0 subsampled jpeg2000 files.
Known issues
Some shell environments may require the use of quotes for each input XML file when using the command line reel frame range option

You can access the new product here:
https://customer.dolby.com/content-crea ... tools-v540
Best regards,
Dolby Laboratories, Inc.

[EncodeThis420]

Hi everyone. I am struggling to understand the below in regards to the hybrid dolby vision profile 8 format. i really hope that someone can help explain this to me:

I keep seeing encodes that take a dolby vision profile 5 stream and combine it with a hdr10 (and/or) hdr10+ stream. I see everyone saying that this helps with compatability. So if you have DV device, it’ll play DV, otherwise it’ll play HDR10/HDR10+.

I have been told that this messes up the original artistic intent of the Dolby Vision Profile 5 stream because it’s using 2 different sources with 2 different gradings. So my question is - if you have a DV device, is the DV profile 5 grading being changed at all from the original? in other words, when it plays back in DV, is it using the HDR10 base layer or is it only using the DV profile 5 info? Hope that makes sense.

I am asking because it seems like ALL of the new DV encodes I’m seeing are in this hybrid format. If it does indeed change the image from the original intent then I have no clue why all of the current encoders are using this format…

If I am wrong with any of the above, please just let me know the correct way to understand it.

Thank you!

[qwertiio]

interesting new options in the new dolby vision tools pack:

https://slow.pics/c/5irgGGKG

L1 Analysis Tuning:
Index Title
0 Legacy CM4 L1
1 Most Highlight Detail/Most Mapping
2 More Highlight Detail/More Mapping
3 Balanced
4 Less Highlight Detail/Less Mapping
5 Least Highlight Detail/Least Mapping
Analysis tuning provides user control over the strength of filtering used during analysis. Six options are
provided that match the options found in partner tools after updating to the 2023 New CM4 L1. As the
index increases, so does the amount of filtering, lowering the L1 max.
Stronger filters may result in clipping of fine details in the highlights while reducing the amount of
mapping. By default, cm_analyze uses the 1 – Most Highlight Detail / Most Mapping
with the intent of preserving as much highlight detail as possible, with the exception of when the --
live-infinite-filter option is invoked. In this case, the analysis defaults to 0 – Legacy CM4
L1 to match the Dolby Live Distribution Processor product

Dolby is please to announce the immediate availability of the Dolby Vision Professional Tools v5.4, available now at customer.dolby.com.

New Feature Summary in this release
New L1 analysis features in cm_analyze
Added 2023 New L1 frame statistics algorithm with user-tuneable spatial filtering, new hybrid mid measurement, unified gamut analysis, and improved black measurement.
New statistics are enabled by default, and tuning can be adjusted using --analysis-tuning <arg>
See the user manual for more information
Older frame statistics are available using --legacy-cm4-l1
Improved file-based long play analysis algorithm
New temporal smoothing algorithm
New backwards compatibility metadata for frame-by-frame analysis
New difference detection to group scenes together
Added --live-infinite-filter to match analysis done in Dolby's Live Distribution Processor
Added new stereoscopic analysis for shot based and file based / frame-by-frame analysis. Right eye view can be added using --input2 <rightEye>
Added new global looks that can be added to your analysis, using --global-trims <globalLook.xml>.
See the user manual for more information
--source-format now requires both the transfer function (pq) and color gamut (p3d65 or bt2020)
• Added new functions in metafier for handling per-shot L5:
--validate performs new checks & reports
looks for and validates per-shot L5. An error will be reported if the per-shot canvas does not match the global canvas
Examines CM 4 XMLs (v4.0.2 & v5.1.0) for "illegal" legacy L1 values.
Separate count for default Target ID 1 L8 trims
--aspect-ratios <arg1> <arg2> now updates canvas of per-shot L5 and removes
See the user manual for more information
--preserve-per-shot-active preserve the current global active aspect ratio as per-shot L5 when updating aspect ratios
--global-trims <globalLook.xml> can inject a global look by applying specified trims to all shots & frames within an existing XML
See the user manual for more information
Cm_offline improvements:
Added new 300 nits P3D65 & BT.2020 targets to cm_offline
Added flag --skip-qt-st2086-min-check to permit cm_offline to continue when a QuickTime MOV's ST.2086 minimum (mdcv) does not match the Dolby Vision master display minimum
Bug fixes in this release
Fixed minor analysis bugs from 5.4-beta-1
Fixed an issue with rejecting 4:2:0 subsampled jpeg2000 files.
Known issues
Some shell environments may require the use of quotes for each input XML file when using the command line reel frame range option

You can access the new product here:
https://customer.dolby.com/content-crea ... tools-v540
Best regards,
Dolby Laboratories, Inc.

I really need this

btw, any eta for this update on the script????

[RESET_9999]

I really need this

btw, any eta for this update on the script????

new beta is available. You need to update your cm_analyze.exe (can be downloaded on dolby website)
L1 tuning can be configured at line 62. Default mode (1) seems to produce the best result.

3-1 can now accept external shot list input which means it's possible to take the shot list from retail RPU and use it with the CM analysis for more accurate metadata.