Overview
For Netflix productions, metadata can be delivered one of 2 ways:
-
Embedded in the Video MXF of a native IMF Package - can be viewed by two methods:
- using clip inspector in supported software (e.g. Colorfront Transkoder)
- extracting the Dolby Vision Metadata via Dolby Metafier
-
Separate XML Sidecar metadata file
- Only acceptable for non-native IMF picture file delivery, which will only be applicable for specific use cases, as agreed upon in advance with your Netflix Post Management/ Asset Management representative.
- Ex. ProRes4444 XQ + DoVi metadata .xml Servicing Turnover
- Ex. Legacy VDM deliveries + DoVi metadata .xml
Key Values
Level 1 (Image character analysis values) - required for ALL shots
- Baseline requirement to generate Dolby Vision metadata.
- Tone mapping will be applied based on the values generated by this default analysis.
Dolby Vision 2.9:
Level 2 <TID 1> (SDR Rec. 709/ 100 nit Trim Pass values) - NOT required
- Will only be included if creative talent decides additional color adjustments need to be made on a scene-by-scene basis.
- If additional Level 2 TID trim passes are planned, the TID 1 (SDR Rec. 709/ 100 nit) trim pass should be completed before any others.
Level 2 <other TIDs> (additional target display Trim Pass values) - NOT required
- Each TID refers directly to a specific target display output.
-
Common TIDs to be included (after the SDR Rec. 709/ 100 nit trim pass is completed) are:
- TID 48 (P3 D65/ 1000 nit or TID 49 (Rec. 2020 D65/ 1000 nit)
- TID 27 (P3 D65/ 600 nit)
Dolby Vision 4.0:
Level 8 <TID 1> (SDR Rec. 709/ 100 nit advanced Trim Pass values) - NOT required
- Will only be included if creative talent decides additional color adjustments need to be made on a scene-by-scene basis.
- If additional Level 8 TID trim passes are planned, the TID 1 (SDR Rec. 709/ 100 nit) trim pass should be completed before any others.
- Backwards compatible DoVi 2.9 L2 trim values will automatically be generated based on the DoVi 4.0 L1, L3, and L8 values
Common QC Issues
Software Incompatibility
Issue: Some versions of certain mastering platforms may be unable to register/ read DoVi Metadata generated by other color grading/ mastering platforms.
Best Practice: If you experience issues applying the DoVi metadata to output an SDR derived version using your prefered mastering tool, extract/ validate the Dolby Vision metadata XML via Dolby Metafier, and open the extracted .xml with a text editor of choice. Use Dolby CM Offline to render an SDR derived standalone master from the DoVi metadata to complete the SDR QC pass.
Image Character values
Issue: The metadata incorrectly contains <ImageCharacter> values of 0,0,0 for any shots that are not entirely black frames.Only entirely black frames would have <ImageCharacter> values of 0,0,0.
Best Practice: Metafier validation includes checks for any L1 0,0,0 values for shots that are longer than 1 second in duration. Confirm if the shots with <ImageCharacter> values of 0,0,0 contain image data (are not black frames) and that the resulting analysis was not intended. Re-analyze, and fix.
Color mapping/ timing error
Issue: When shot durations in Dolby Vision metadata do not align with cuts in the video, there will be color/ luminance mistimes (shifts) midshot in the SDR derived output.
Best Practice: Extract the metadata and validate for overlapping and missing durations using Dolby Metafier prior to QC. Also, use clip inspector in supported software to check for metadata value changes that do not align with cuts. This always requires a fix.
Canvas and Image Aspect Ratio settings
Issue: If either or both of these values are set incorrectly, color mapping will not be applied to the image correctly. If positive lift values are also utilized in the trim pass, this can manifest as shifting / elevated black levels in the mattes for letterboxed/ pillarboxed content, visible in the SDR output derived from the Dolby Vision metadata. If the active image area is larger than what is represented in the metadata, it will be visible as “translucent” bars in the image where the color mapping is not being applied in the SDR output.
Best Practice: During the Dolby Vision analysis to generate the metadata, ensure that your <CanvasAspectRatio> is set to the full resolution of the frame (represented as a non-negative, decimal number).
-
For UHD (3840x2160) or HD (1920x1080) deliveries - this should always be 1.77778
- Per the latest Netflix delivery specifications, UHD and HD deliveries are the only resolutions supported for Native IMF delivery.
-
For full 4K deliveries (4096x2160) - this should always be 1.89630
- Will only be accepted, if previously agreed upon with your Netflix Post Management/ Asset Management representative.
Ensure that your <ImageAspectRatio> is set to the A/R of the active image within the frame, minus any mattes (represented as a non-negative, decimal number). If the aspect ratio is a continuous decimal number, a maximum of 5 decimal places of precision is required to account for one-line discrepancies. If the aspect ratio is a whole number integer or a finite decimal number, it can be represented with the amount of decimal places to match its value from a single decimal place up to 5 decimal places of precision.
- For a show delivering 2:1 - this can be represented as 2.0 or 2.00 or 2.00000
- For a show delivering full frame - this should match the Canvas Aspect Ratio
If either or both of these settings do not match the video track as delivered, it will always require a fix.
Additional notes: If there are multiple creatively intended active image aspect ratios, <ImageAspectRatio> should be set to the largest active image aspect ratio that is utilized. Then, include L5 dynamic aspect ratio metadata values for the shots that deviate from this global Image Aspect Ratio.
- Filmlight Baselight v5.3 and up now supports the addition of L5 dynamic aspect ratio metadata.
- For titles finishing in previous versions of Baselight, the addition of L5 metadata is not supported, the workaround is to set the global <ImageAspectRatio> to the largest active image aspect ratio that is utilized. Then, avoid any positive lift values in the trim pass to ensure shifting/ elevated black levels do not occur in the mattes in the SDR derived output during those shots that deviate from the global A/R.
- For shots that are BOTH letterboxed AND side-matted/ pillarboxed (i.e. both X and Y active image dimensions are less than the full canvas resolution), Dolby does not currently support accurately capturing and reading L5 metadata values to account for both X and Y coordinates. While some color grading/ mastering platforms will accurately identify the correct active image, they may not save these values within the DoVi L5 metadata in a way that can be interpreted to correctly apply color mapping, ignoring the mattes both top/ bottom and left/ right.
- L5 metadata, if supported by the color grading/ mastering platform, should be applied based on X (horizontal) OR Y (vertical) active image aspect ratio, but not both.
- Positive lift values should be avoided in the trim pass to ensure shifting/ elevated black levels do not occur in the mattes.
- Starting in Dolby Metafier v5.0.1 there is an additional option to include a check for all shots containing positive lift values during validation.
For QC partners: Please make sure you are always verifying both the Canvas and Image Aspect Ratio settings in the Dolby Vision metadata are accurate to the full resolution and active image aspect ratio of the video track. If you are unable to verify the metadata, in the case of Cast QC where only the HDR/ SDR encoded streams are served, be mindful of visual errors that would indicate there is a problem with these settings. If there is shifting/ elevated black levels in the mattes that are visible only in the SDR stream, that would indicate a potential issue with the Canvas or Image Aspect Ratio settings. Also, if there are “translucent” bars in the image top/ bottom or left/ right that are visible only in the SDR stream, where the HDR color is essentially being passed through unmapped, that would also indicate a potential issue with these settings.
CHANGE LOG
06/23/2022
- Removed language around a "challenging workaround" for L5 metadata in Filmlight Baselight V5.3+ and the recommendation to "reach out to your Production Technologist for guidance". Per feedback from CT, this was misleading, and Technologists, in general, were not aware of any said workaround.
05/25/2022
- Updated L5 statement for Filmlight Baselight
1/8/21
- Updated Image Character 0,0,0 issue section with improved wording
v 6/3/2020
- Changed references from Dolby CM Offline to Dolby Metafier throughout for metadata extraction/ validation.
- Added clarification and examples to address when the separate XML sidecar is an acceptable delivery method for Dolby Vision metadata.
- Added additional target display information to Key values section.
- Added distinction between Dolby Vision 2.9 and Dolby Vision 4.0 trim pass values to Key values section.
- Changed “Software conflict” section to “Software Incompatibility” and removed specific references to Colorfront Transkoder and Rohde & Schwarz Clipster versions, in favor of more generic guidelines around potential compatibility issues between different mastering platforms.
- Added Dolby CM Offline render information to the Best Practices under the Software Incompatibility section.
- Updated Best Practice under the Image Character values section.
- Added guidance around shots that are both letterboxed and pillarboxed to Additional notes under the Canvas and Image Aspect Ratio settings section.
- Added clarification around accepted UHD/ HD vs. 4K deliveries to Best Practice under Canvas and Image Aspect Ratio settings section.
- Added information around the additional option to check for positive lift values during metadata validation when using Dolby Metafier v5.0.1(or later).
- Removed links to translations for previous version of this article, which are now outdated.
Translations