2.1.2

libdovi changes

• rpu/parser: Performance improvements
  • Use of ArrayVec to reduce heap allocations.
  • Removed redundant clone when encoding RPU to bytes.
  • Use of larger LUT for CRC32 calculation, ~15% speed up.
• av1: Add dovi_parse_itu_t35_dovi_metadata_obu to C-API by @galad87 in #301
• rpu/parser: Parse ext_mapping_idc from el_bit_depth_minus8 by @quietvoid in #264
  • Adds support for profile 20 sample.

Full Changelog: 2.1.1...2.1.2

Download dovi_tool 2.1.2

File	Platform	Checksum
dovi_tool-2.1.2-universal-macOS.zip	Apple Universal macOS	checksum
dovi_tool-2.1.2-x86_64-pc-windows-msvc.zip	x64 Windows	checksum
dovi_tool-2.1.2-aarch64-pc-windows-msvc.zip	ARM64 Windows	checksum
dovi_tool-2.1.2-x86_64-unknown-linux-musl.tar.gz	x64 Linux	checksum
dovi_tool-2.1.2-aarch64-unknown-linux-musl.tar.gz	ARM64 Linux	checksum

2.1.1

Bug fixes

• Fixes in RPU generation from XML:
  • Supports decimals in level 6 MaxCLL/MaxFALL metadata without erroring.
  • Properly sets source_min_pq and source_max_pq from L0 mastering display.
• Editing min_pq and max_pq could be overridden by level 6 metadata when set to 0.

 

libdovi changes

• av1: parse/write EMDF wrapper properly by @quietvoid in #285
• av1: Discard trailing zero bytes in EMDF payload by @quietvoid in #293
• rpu/parser: Refactor handling of trailing bytes by @quietvoid in #294

Full Changelog: 2.1.0...2.1.1

Download dovi_tool 2.1.1

File	Platform	Checksum
dovi_tool-2.1.1-universal-macOS.zip	Apple Universal macOS	checksum
dovi_tool-2.1.1-x86_64-pc-windows-msvc.zip	x64 Windows	checksum
dovi_tool-2.1.1-x86_64-unknown-linux-musl.tar.gz	x64 Linux	checksum

2.1.0

Bug fixes

• It is now possible to read RPU files that are over 250M in size. (#233, fixes #221)
• Fixed RPU generation from HDR10+ generated by Resolve when the timeline starts at 86400 seconds.

Features

• Improved export subcommand to allow specifying what data to export. (#256, implements #223)
  • -d, --data: List of key-value export parameters formatted as key=output,key2...
    • all - Exports the list of RPUs as a JSON file
    • scenes - Exports the frame indices at which scene_refresh_flag is set to 1
    • level5 - Exports the video's L5 metadata in the form of an editor config JSON
  • See README section for details and examples.
• Added remove subcommand, which outputs only the base layer HEVC file. (#257, fixes #232)

 
Full Changelog: 2.0.3...2.1.0

2.0.3

Bug fixes

• Fixed error when RPU is missing L1 metadata for some frames, using info or plot. #216
  • In those cases, a default L1 is used to avoid gaps in the plot.

Features

• Added --hdr10plus-peak-source to generate subcommand. #217
  • Now HDR10+ generation defaults to histogram, the old behaviour is now max-scl.
  • See readme for more info

2.0.2

Bug fixes

• Fixed rare old bug where writing HEVC streams could contain a non conformant RPU NAL.
  • It causes an error when extracting again from affected HEVC streams.
  • The solution is to recreate the file with version 2.0.2.
  • All processed (and RPU parsed) HEVC files could be affected, as long as they were not remuxed as remuxers would correct the stream. So this bug essentially only affected raw bitstreams.

2.0.1

Bug fixes

• Fixed plot L6 metadata line when there are more than 2 different L6 blocks. (#208)
  • The line now limits to 2, with a count for the actual total of different L6 metadata combinations.

2.0.0

Bug fixes

• Fixed incorrect MMR coefficients (#197).
  • Users of the mmr_coef{_int} array are the only ones affected by this bug.
• Fixed error with --drop-hdr10plus for particular edge cases.

Features

• Refactored the RPU mapping structure. The changes also affect the exported serialized JSON.

  • This changes the structure of rpu_data_mapping. The NLQ parameters are also moved into it.

• Extended info --summary output:

  • RPU mastering display: The mastering display recorded in the RPU, from source_{min, max}_pq.
  • RPU content light level (L1): The MaxCLL and MaxFALL values calculated from L1.
  • L6 metadata: The L6 metadata present in the RPU.
  • L2 trims: The target display max brightness for which the RPU has L2 trim controls.

• Changed mode 2 behaviour to automatically enable remove_mapping for profile 7 FEL only.

  • This improves the default behaviour so that the output image is correct on playback.
  • For the old behaviour, mode 5 was added.

• Added plot subcommand, to plot the L1 metadata into a graph. It also shows the same info as info --summary.

• Improved memory usage, performance. RPU operations should be noticeably faster, ~50% speed up.

Notes:

• Linux release built on 893d683.

1.6.0

Bug fixes

• Fixed injecting RPU when base layer contains SEI_SUFFIX NALUs.
• The editor subcommand now errors if it encounters invalid fields in the JSON.

Features

• Improved memory usage by ~13%, performance by ~21%. Allocations reduced by ~50%.

Notes

• There is now a macOS universal binary release (#178).

1.5.8

Features

• Added source_rpu and rpu_levels to edit config. #188
  • Allows copying metadata from one RPU to the other.
• Added l1_avg_pq_cm_version to generate config.
  • Allows overriding the minimum L1 avg_pq CM version.
  • Example use case: Some grades are done in CM v4.0 but distributed as CM v2.9 RPU.

1.5.7

Bug fixes

• Fixed building from code without a git repository.
• Fixed parsing HEVC files containing filler data NALUs.

Features

• Improved the clamping for L1 metadata when generating.
  • For CMv2.9, the average luminance has a minimum of 2.5 nits (819/4095).
  • For CMv4.0, the minimum average is 10 nits (1229/4095).
• Added support to drop HDR10+ when it exists in a SEI NALU containing multiple messages.

Notes

• The clap library used for the CLI argument parsing was updated to version 4.0.4.
  • The only user facing change should be that there is no more coloured help text, it now makes use of bold and underlined text.