1.4.7

Bug fixes

• Properly parse the actual fields in profile 7 RPUs. #133 contributed by @saindriches. Fixes #131.

Features

• Add partial support for compressed RPU format.
• Improved RPU extraction from invalid files where multiple RPUs existed for a single frame. Fixes #129.

Notes

The release workflow has been updated to improve the artifacts and make them more convenient.
Now, the archives only contain the executable without nesting it in a dist directory.

Also, Windows and macOS releases were switched to zip archives.

1.4.6

Bug fixes

• Allowed noop conversion when converting a profile 8 RPU with mode 2. Implements #112.
• Fixed parsing when the RPU NALU has multiple trailing 0 bytes. Fixes #122.

1.4.5

Bug fixes

• Fixed generate CLI argument conflicts, regression after migration to clap v3.
• Corrected L9/L10 primaries presets for generation from XML.

1.4.4

Migrated to clap v3, replacing structopt for handling CLI arguments.

Bug fixes

• The dolby_vision lib now handles variable length metadata blocks correctly.
  • This can be used by L8, L9, and L10 metadata levels.
  • Implements fixes for #104 and #105.
  • Thanks to @saindriches for figuring out most of it.
• When generating a CM v4.0 RPU, a default L9 metadata block is added to conform to Dolby spec.
• It is now possible to inject RPUs into a video file that already contains Dolby Vision RPUs. Fixes #108.
  • In that case, the existing RPUs are ignored and replaced.
• Corrected L3 metadata values from XML.

Features

• RPU generation from XML now supports L10 and L11 metadata. As well as fixes for variable length L8 and L9.
  • L10 metadata is derived from the target displays in the metadata, and can support custom primaries.
• Added a way to edit the existing L9 metadata in RPUs. Only supports preset primaries, refer to documentation.

1.4.3

Moved the documentation to a separate docs directory.
Added more examples for use with generate subcommand.

Bug fixes

• Issue #91: Fixed regression to allow users adding L2 metadata when generating.
• Issue #97 (fix by @nekno): Improved output file name when using convert subcommand to reflect the presence of EL/RPU better.
• Improved generator config JSON default values to not error unnecessarily.

Features

• Added the possibility to specify any default metadata block levels to be used for the whole sequence when generating RPUs.
• Added a way to edit per-shot/per-frame metadata even when generating from HDR10+ or madVR measurement files.

1.4.2

Bug fixes

• Fix regression when the DM data has no blocks, #87

1.4.1

Bug fixes:

• Fix #86, where dovi_tool would reorder metadata blocks when parsing, resulting in an incorrect CRC32 compared to the original data.

1.4.0

Big feature/breaking release to implement support for DM v2/CM v4.x metadata.
Now properly parses/writes metadata blocks for both display management metadata versions.

Supported metadata levels:

• DM v1 (CM v2.9): L1, L2, L4, L5, L6
• DM v2 (CM v4.x): L3, L8, L9, L10, L11, L254

Important changes:

Generator

• Implemented proper support for L3, L8 and L9 metadata for RPU generation from XML.
• Improved the template to allow more flexibility. Now allows specifying shots (with per-shot or per-frame metadata blocks).
• Refer to documentation and example.

Editor

• Added convert_to_cmv4 option to convert an existing RPU to DM v2, defaulting to adding L11 metadata.
• It is now possible to remove polynomial/MMR mapping from the RPU using remove_mapping
• Refer to documentation and examples.
• For L11, see convert_to_cmv4 example.

Exported JSON from info or export now have DM metadata split in two objects:

• cmv29_metadata and cmv40_metadata, which have the same structure: a number of blocks, and list of blocks.

 

For the C API, see changelog.

1.3.2

Bug fixes:

• Fix regression in conversion from profile 5 to 8.1. Fixes #80.
• Fixed incorrect level 3/4 metadata block lengths. Fixes #81.

1.3.1

Updated dolby_vision lib to 1.3.1.
Mostly changes the default initialized list of values in the parsed RPU.

Added --el-only flag to demux subcommand to only output the EL video stream.

Edit: Regression in mode=3 conversion.