Dolby Vision metadata is a critical part of the overall Dolby Vision system. It is required because consumer displays do not all have the same display characteristics, particularly in terms of peak brightness, contrast range and color gamut support. Without Dolby Vision’s metadata and standardised tone mapping processes it would be up to each display and each manufacturer to decide how to modify the content to fit within the individual display characteristics. 

Dolby Vision metadata enables the content creator to previsualize and control this mapping. In systems without a standardised tone mapping process like HDR10, there can be a large variation in the methods and mappings used, even between consumer devices that have the same or similar display characteristics.

Dolby Vision metadata is dynamic metadata which means it can change many times throughout the content. The decisions taken for one shot do not have to be the same as those taken for a different shot. At its core, Dolby Vision metadata is carrying information about the brightest pixels, darkest pixels, and the average picture level, but there are a range of additional parameters that can be used for more detailed control.  Dolby Vision metadata can be created automatically, or can be a combination of automatic and manually generated information.  In the latter case a colorist can create multiple different trim passes where the manually configured metadata individually adapts each shot to different target displays.  

There are two methods for creating Dolby Vision metadata, these are shot analysis mode and long play analysis mode. Which mode to use will depend on available production time, workflow, and whether manual control of trim passes is required. 

While not exclusively the case, if you are delivering a mezzanine asset for delivery to an OTT service it is likely that you will use shot-by-shot metadata (shot analysis mode) and optionally create manual trim metadata for more direct control over the playback display mapping process. 

If you are delivering an encoded Dolby Vision file for upload to a video sharing service, it is likely that you will deliver frame-by-frame metadata (long-play analysis mode) and this metadata will be created automatically as part of the file export/encoding process.
 

Shot-by-shot metadata (shot analysis mode)

This is the mode that provides the content creator with direct control over the consumer replay on different displays. Shot analysis mode analyzes each shot in a timeline, and then the colorist can manually apply trim controls to tailor the mapping to a range of different targets. You should have a color grading system that supports Dolby Vision shot analysis mode and an active Dolby Vision license is required on the color grading application in order to (optionaly) trim the metadata for selected shots. For an example of how to run the automatic shot analysis see this article and for an overview of how to use trim passes see this link .

With shot analysis mode it is important that this metadata is generated as part of the content creation/color grading process, as particularly if trim metadata is being created this needs to be done by a trained professional with access to a suitable grading monitor. Fortunately, this functionality is available in many major color grading applications at a wide range of price points. See this link for a list of supported applications.
 

Frame-by-frame metadata (long-play analysis mode)

Long play analysis mode is an automatic process for creating metadata where a video file, capture or HDR master is analyzed and Dolby Vision metadata is created automatically for every frame of the video master. It is recommended for whole-title analysis of HDR assets that have not been mastered in Dolby Vision. It can also be used if a film was restored or converted to HDR to quickly make it a Dolby Vision deliverable. Longplay is only for file-based metadata generation as it looks 20 seconds before and after the current frame, regardless of frame rate. The default analysis tuning for all long play analysis is "Most Details / Most Mapping" when no user interaction is expected to validate the results or should follow the user project selection (default-Balanced) when used within the bounds of a shot.

As this is a completely automatic process, it can be run at various points in the workflow, even as late as in the final distribution encoder. It analyses each individual frame of content and the metadata can change on even a frame-by-frame basis if that's appropriate given the content. With long play analysis mode it is not possible to manually trim the metadata, but it is a great option where production time is limited or where there isn’t a requirement for such detailed control over the tone mapping process. Read more about using long play anaysis using the Dolby Vision Professional Tools in this article.
  
Long-play analysis is a great way to create Dolby Vision mezzanine deliveries for OTT services from existing HDR assets. However, if you are planning to deliver to a video sharing platform as Vimeo, you can export directly in Dolby Vision profile 8.4 that will calculate the Dolby Vision metadata during the export process. Dolby Vision profile 8.4 is supported on these Dolby Vision desktop encoders .