Future File Formats for Media

June 18, 2013, Entertainment Technology in the Internet Age, SPMTE and SCIEN conference, Stanford, CA—A panel took on the topic of future file formats and the trends and implications for Web distribution. Howard Lukk from Walt Disney Studios chaired the presentations. Speakers were Pierre Lemiuex from Sandflow consulting and John Hurst from Cinecert.

Lukk opened by describing the problems with distribution, servicing, and mastering content. The biggest problem is the multiple different versions of the final assets. There are various aspect ratios for theatrical and home use and many localizations for language and culture. A typical release has 26 language versions, but can be up to 42, each language has to be in 3 aspect ratios and 3 formats. This amounts to 378 versions for any video, not including other versions for Web distribution.
In addition, some of the front-end and end credits are changed for different locations, as well as some internal content. A joke in English for the USA needs to be changed for the UK. This supplementary localization is on top of the other variations. Total possible variations can be over 35,000.
The standards for file format requirements is IMF. This interoperable master format describes the master format and scripts for all versions, and only the changes are stored. The image tracks are mostly constant while the underlying language changes. In some cases, they replace the commercial leader black areas with pointers.

Lemieux provided an IMF overview. The source content is mapped to the distribution channels and defines the deliverables. The master package is based on a digital cinema standard and XML. The composition play list (CPL) and all associated assets are merged to synchronize all essence: audio, video, captions, etc.
The assets are moved into a timeline where segments and sequences of audio and video are put together. The audio is in 24-bit PCM at either 48 or 96 kb/s. subtitles are in SMPTE TT format, and the images are in various file formats like JPEG 2000, MPEG 4, etc.
The IMF is extensible over time. The output profile lists identify the transforms. A packing list in XML identifies all the files needed for the film. Late bindings and other supplements are added at runtime for the process. An asset map packages the totality for transfer to physical media. There are specifications and standards for most of the functions.
Hurst added more IMF details. The XML base calls various codecs through a material exchange format (MXF). Key length value packets are concatenated into strings. The MXL file format has a header with metadata and the image track files. Sound track clips are wrapped into the video sequences through a text track file.
A UL resolver uses basis encoding rules (BER) for integer decode. All of the basic technologies are open source and are available.

Distribution of content over the Web? How to handle full distribution to all channels?
Lemieux commented that the transformations have a direct impact. Source content for the Web in one of the deliverables, but captions and other lower-level assets may suffer from the conversions.

Storing files and versions?
Lukk noted that in-country people edit for localization. Some functions are not directly structured.
Hurst added that they use AS-02 to track files and changes in version files. Everything is based on a digital cinema standard.