You may also be interested in:

Delivery specifications are produced by broadcasters and streaming companies to describe how contributing production studios should format and submit their assets for deployment. These are detailed documents and they cover all of the supporting assets that go with the program.

In order for the program to appear in the viewers dashboard, metadata needs to describe various properties of the program. Supporting text, graphics, posters, trailers and featurettes must all be present.

The Netflix delivery specification is a prime example of the level of fine-detail required to specify compliant delivery of programs from external producers. There are equivalent documents provided by all broadcast and streaming companies. They cover broadly similar topics with some details altered to suit their own ingest workflows.

Let’s step back from a specific case such as Netflix and look at the more generalized common requirements across a range of deployment platforms.

Let’s pick up where we left Part 1. 

Audio Assets

The audio will most likely be embedded into the video assets. Additional audio may need to be supplied for accessibility and localization of the program.

Audio content is defined with the following properties. This is an aggregated list compiled from various delivery specifications. The target delivery specification may constrain them to a smaller sub-set: 

Audio Sample Rate & Size 

Generally acceptable sample rates are:

The 2 and 4 times super-sampled quality are designed to allow mixing and other audio processing without introducing artefacts.

The samples are captured at 16 or 24-bit resolution.

Audio Channel Configuration

The target platform may be capable of delivering audio in several alternative viewer selected presentation formats. If necessary, add support for these formats to the manifest:

• Mono audio (Rarely needed).
• 2 channel Stereo.
• 5.1 surround-sound.
• 7.1 surround-sound.
• Dolby Atmos.

If audio is embedded into the video content file, the metadata should flag that it is replaced with the stand-alone audio files.

Surround-sound Channel Mapping

Surround-sound may be delivered as a single multi-track file or with each channel in a separate file. Multi-track 5.1 surround-sound may have the tracks arranged in one of four different configurations.  The convention for 7.1 is simpler with only one mapping defined.

The channel mapping order is important and must conform to the delivery specification. There are conventions but these are sometimes altered.

This is a set of conventional labels used for tagging audio channels. The track number assignments for 5.1 and 7.1 surround-sound shows the channels used for those configurations:

Surround-sound Down-mixing

Apple TV+ supports down-mixing to other formats from a multi-track Dolby Atmos original carried in a separate file:

• Stereo left and right 2.0 format.
• Surround-sound 5.1 format.
• Surround-sound 7.1 format.

Stereo down-mixing should use appropriate combinations of the left and right channels:

Audio Stems With Alternative Tracks

Supplementary audio tracks are delivered in separate stems (tracks). This facilitates localization, placement of the audio in a 3D soundscape and accessibility support.

The BBC content delivery book goes into great detail about how the audio content should be split into multiple stems:

• Audio description.
• Dialog without music.
• Multiple spoken foreign language dialog.
• Cast & crew commentaries.

These should all be synchronized to the video so that dynamic mixing and user preference settings can switch stems on and off as needed.

Background Loops

Background audio loops can be played in the user interface while the viewer is browsing the video description and looking at the stills gallery. These are used if the background is a still image rather than a moving background video.

Audio Delivery Format

There are a variety of ways to embed audio or deliver it as separate files. Embedded audio should remain synchronized to the video without the player needing to do anything complicated with start-time offsets.

Audio Rights Clearances

If there is any music playing in the background of your scenes, the rights to use it should be cleared beforehand.

Rights are cleared through these organizations:

• PRS for Music
• PPL - Phonographic Performance Limited
• MCPS - Music Publishers Association clearance is handled through PRS

Loudness Issues

Accurate loudness estimation is important. Viewers are annoyed if the sound levels change during advertising breaks or when changing channels.

The loudness specification is described in ATSC A/85 and ITU-R BS.1770-3. The preferred loudness recommendation is:

-24LKFS (+/- 2 dB)

This measurement uses Loudness, K-weighted, relative to Full Scale (LKFS). It defines the perceived loudness. LKFS is used in the USA. European broadcasters use Loudness Units relative to Full Scale (LUFS) measurements.

Dolby has created a freely available tool for assessing the sound level. This concentrates on passages of dialog since everything else is mixed around that. Register to download the Dolby Speech-Gating software if you need it.

Audible Profanity

Broadcasters prefer to bleep out or include profanity themselves according their own taste and decency guidelines. Streaming platforms tend to be less restrictive in this respect since they are not governed by the same guidelines as on-air broadcasters. This is a loophole in current safeguarding regulations.

A file containing time-coded references to potentially offensive audible items should be provided in the manifest.

Ofcom publishes a list of proscribed words for radio announcers. These should never be uttered on-air. This is a good starting point for compiling your own list. Any kind of abusive language or hate speech is treading on dangerous ground if it is included without justified reasons for its use.

Imagery

A variety of graphical image assets may be required for use when building user interface dashboards or as promotional assets in sites such as IMDB:

There are rights processing organizations who specialize in arranging clearance for images and visual works. They are different to the literary and audiovisual licensing organizations.

Alpha-channel Masking

If the image format supports alpha-channel mapping then masking should be provided. This is particularly applicable to PNG graphics files. Apple TV+ uses this to present the cast and crew portraits in circular frames.

Animated Content

Historically Adobe After Effects® has been used to create animated segments. Alternative software in current use would include projects created in DaVinci Resolve and Apple Motion.

The finished program must be delivered in a flattened and rendered format. Some clients require the animation assets collection to be delivered as well. This facilitates later revisions.

The BBC guidelines are very specific about how to prepare an After Effects project for delivery. Similar preparation can be applied to DaVinci and Apple project containers.

Text Layers

If there are burned in forced narrative texts, you may need to provide a textless version of the video so these can be localized for other languages. A better approach is to not burn them in at all but provide a separate VTT track so they can be shown on-screen even when subtitles are turned off.

Typography & Fonts

List all of the fonts used in any on-screen rendered typography. If you use proprietary fonts, you may need to license them separately for use in video.

Some fonts are optimized for screen use. A font that may look good in a hard-copy printed form might not render attractively on-screen.

Do not deliver the original font suitcases in the manifest. The license to use them does not allow redistribution. The client must purchase their own copy.

Subtitles & Timed Text

Subtitles should be localized and delivered as VTT files. These can be described in the metadata if there are multiple files for different languages.

In the USA, these are called Closed Captions. There may be specific formatting requirements for these and how they are described in the metadata.

Timed text tracks can also be embedded directly into the video file. QuickTime labels this as a CC track. MXF files use a ST 436 (part 1 and 2) Vertical Ancillary Data (VANC) track. Some target players are unable to read these embedded timed text tracks. The separately supplied VTT files approach is a recommended and preferred solution in most cases.

Chapter Marks

VTT files can also be used to create chapter marks. These may also require images for each chapter mark so they can be presented as a row of selectable scenes. The VTT file would contain a reference to each image and a descriptive title for the chapter. The end-credits would also be tagged as a separate chapter as would the starting credits/intro. Recaps and trails should also be marked.

Text Blocks

Large blocks of editorial text are useful as lengthy descriptions of program content and plots. Director, producer, cast and crew biographies and extensive information texts are presented by the Apple TV+ user interface.

Text could be delivered in any of these formats depending on the delivery specification:

• HTML styled with CSS.
• Unstyled HTML.
• XML needing to be rendered.
• Raw plain text.
• Raw plain text with Markdown annotation.
• Rich Text Format (RTF) files.
• MS Word Doc files.
• MS Word DocX files.

You may need to get clearance for literary works that are read out aloud or included in the visual presentation. The CLA organization licenses and collects royalties which are then redistributed to publishers via PLS and individual authors via ALCS.

Container Packaging

The assets are delivered together in a package. This might be loosely collected into a folder or encapsulated into an archive file (tar or zip files for example).

Here are some common packaging techniques used by Netflix and others in their delivery specifications:

The AS-11 file format specifications are all available from the AMWA web site.

Photo Sensitive Epilepsy (PSE) Tests

Use a photosensitive epilepsy (PSE) test tool to examine the video for excessive flashing. If necessary tag the metadata or put a health warning slate on the front of the content. A metadata tag should trigger a warning interstitial created by the player when the content is viewed.

Consult the ITU Recommendation ITU-R BT.1702-3 which describes “Guidance for the reduction of photosensitive epileptic seizures caused by television”.

Potential Production Recalls

Avoid costly production recalls where content needs to be re-worked due to compliance or editorial fails that are detected late in the lifecycle or after the broadcast.

These are examples of compliance issues that need to be addressed with urgency.

Checking for these before delivery will save time, money and trouble and cause less damage to your reputation:

• OFCOM complaints.
• Music not properly cleared for use.
• Archival footage not properly cleared for use.
• Viewer sensitivities and safeguarding.
• Unnecessary contraventions of taste and decency guidelines.
• Uncleared telephone numbers and other privacy violations.
• Faked content.

These are examples of potential editorial mistakes that need to be fixed. There will be others not listed here:

• Poorly executed visual effects compositions.
• Bad edits.
• Out of gamut (illegal) colors.
• Codec burst errors when encoded.
• Compression artefacts (blocking).
• Broken lip-sync.
• Visible production kit (boom arms and second cameras in shot).
• Credit changes (contributor names misspelled).
• Burned in text errors.
• Embedded metadata inaccurate or misspelled.
• Audio tracks assigned to incorrect channels.

Quality Checking

The asset manifest provides a checklist for the production automation. When every asset is available, the project status can be set to ‘ready’ and flagged for sign-off. Once the necessary Quality-Assurance checks are complete, dispatch to the client can be initiated. These items are key:

• All enumerated items listed in the manifest should be present.
• All essence items should conform to the delivery specification.
• Photosensitive Epilepsy tests should have been run and the content annotated appropriately.
• All potential production recalls should have been checked for and eliminated.
• Loudness protocols specified in the delivery spec should be checked.

Workflow automation saves time but human moderation and inspection should not be entirely eliminated.

Conclusion

Delivery standards all need to do the same job. They are broadly similar in content and scope. There are some differences at the detail level and they may not all require the same manifest of assets to be supplied.

Forward planning to prepare a super-set of the assets equips you to redeploy your production to other platforms later with the minimum of additional production effort.

Appendix - Sourced Delivery Specifications

These are the delivery specifications referred to as source material:

Appendix - Filename Extensions & Formats

This is a non-exhaustive list of file types that might be present in a manifest being constructed according to the guidance in a delivery specifications.

Appendix - Related Standards

These are some relevant standards for contributing professionally produced content. Where there are errata or amendments, their date is taken to indicate the vintage. Some vintages are as quoted in the references in other documents. The older vintage is preserved in case later versions introduce incompatibilities:

Note that there are internal normative and informative references in some standards documents and delivery specifications that describe older SMPTE standards such as ST 436M. The letter ‘M’ suffix is obsolete and these standards have been re-issued with a numeric suffix denoting one of several parts. ST 436M has been replaced by ST 436-1 and ST 436-2 which enhance the specification to describe the encapsulation of VANC data in MXF files. The older ST 436M file only addressed television applications.