AAF for DVD: Enabling a New Era of Interoperability?
Philip De Lancie
January 2000 |
There's no shortage of file formats for the video, audio, graphics, and text that go into a DVD-Video title. But the specific formats supported by the capture, processing, and authoring tools used for DVD vary by vendor, and the conversion of media from one format to another throughout the process can be both an impediment to collaboration and a waste of energy. Further, these formats make little provision for passing anything beyond the content data itself, so authoring tools have no access to information about the editing or processing steps a file may have been through on its way to authoring. Content generally arrives at authoring stripped of its history, already fixed in a form that makes it difficult to change without heading back upstream.
DVD-Video authoring is not unique in facing these hurdles to efficient workflow; the same challenges are faced in areas such as video post, audio production, 3D animation, and interactive multimedia. To address these issues, a group of digital content-tool vendors got together in March 1998 to work up a new file interchange standard dubbed "Advanced Authoring Format" (AAF). Championed by Avid Technology and Microsoft, AAF is also being promoted by Adobe Systems, Pinnacle Systems, Matrox Electronic Systems, the Sonic Foundry, Sony Corporation, and Discreet (Autodesk).
At September's IBC conference in Amsterdam (Europe's equivalent to the NAB show), AAF's backers announced the formation of the AAF Association, "an independent trade organization aimed at promoting the adoption of AAF and directing its future development." The group has also readied an AAF Software Developers Kit (SDK) for MacOS, Windows NT, and SGI Irix (UNIX), available for download from the association's Web site (http://www.AAFassociation.org).
IBC also saw pledges of support for AAF by Quantel, Philips, and Sonic Solutions. With an SDK ready and industry support growing, this seems like an opportune time to take a closer look at what AAF has to offer, and how it might impact the DVD-Video title preparation process.
The fundamental purpose of AAF is to facilitate interoperability between applications used in digital content creation. Unlike QuickTime, AAF is not intended to serve as a delivery format, but rather is exclusively designed to serve the needs of the production community. The idea is to extend the same non-linear flexibility that is currently available within individual applications to the overall creative process.
"An open-file interchange format is essential to enable the all-digital production facilities of the future," says Ed McDermid, Avid's AAF marketing manager. "The effective integration of digital content creation products from multiple vendors has proven to be a significant obstacle to date. AAF represents a broad industry initiative, supported by a consortium of competing vendors, to remove those obstacles."
Once implemented in content-creation applications, AAF is supposed to enable seamless exchange of digital media and metadata between production tools across multiple computing platforms. According to the Association, the significant benefit that AAF provides to end users is the assurance that "compositions" output by AAF-compliant applications will be "accessible by the right tool for the job, without risk of being 'stranded' by proprietary file format restrictions." McDermid adds that the ability to "exchange media, compositions, and other data quickly and efficiently between AAF-compliant systems will improve workflow, save time, and increase creative freedom."
AAF hasn't been invented from scratch. The format builds on Avid's Open Media Framework Interface (OMFI), and is in fact considered by Avid to be OMFI version 3. However, AAF is implemented in an object-oriented C++ environment, incorporates a vendor-neutral plug-in architecture, supports a broader range of media formats and temporal data, and allows the integration of 2D, 3D, text, HTML, and XML objects. According to the Association, bi-directional OMFI/AAF conversion tools are available as part of the SDK.
AAF also uses a different container format than OMFI. Contributed by Microsoft, this "structured storage" container uses a "file system within a file" architecture. The platform-independent file format is able to store a variety of raw media file formats, as well as the complex metadata that describes the usage of the media data.
MEDIA VS. META
The distinction between media data and metadata is a fundamental concept of AAF. Media data (also referred to as "essence data") is the actual content, such as video, audio, graphics, animation, and text. AAF does not impose a universal file format for storing media data. It has some commonly used formats built in, and also provides an extension framework for new or proprietary formats. Applications may use AAF as their native file format, and may store application-specific data in an AAF file.
The format used for media data that has been encapsulated in an AAF file is identified to other applications, making it unnecessary to provide a separate identification mechanism. On the other hand, an AAF file need not actually contain essence data at all, because metadata may be used to reference external media data files.
Metadata is data that describes, performs an operation on, or provides supplementary information about essence data. For example, information describing the format and duration of a media file is metadata. But AAF supports much more sophisticated types of metadata, including what the specification calls "compositional information," which describes how sections of media data are combined and modified. Provision is made for storing information such as in and out points, volume, panning, and time and frame markers. [For more discussion of metadata and metatags, see Robert J. Boeri and Martin Hensel's July 1999 INFORMATION INSIDER column, "Next-Generation Searching: Looking for the Right Stuff"--Ed.]
AAF also supports "version control," allowing an AAF file's data to be edited and revised while retaining the history of the changes, so that that an older version of the file may be recalled. Another type of metadata is "media derivation," meaning that AAF files will retain information, allowing edited media to be traced back to its original source. In theory, then, AAF metadata will carry from one AAF-compliant environment to another a blueprint of the current state of an edited or authored production, as well as a log of the production history.
This support for extensive metadata, including custom metadata categories that may be created by end-users, is what AAF is all about. "It bridges the gaps between all the steps involved in off-line and on-line production processes," McDermid says. "If you capture and retain metadata all the way through, you don't have to revisit decisions and redo things that you already did earlier." In DVD, for instance, pan/scan decisions made in video post will be passed to the authoring application as metadata rather than having to be re-entered during authoring.
MOBS AND ESSENCE OBJECTS
Metadata and essence data are stored within the AAF file structure as objects. The specification describes four main types of metadata objects (Mobs), each containing metadata that is used in different ways. Physical Source Mobs describe the medium used to generate essence data (i.e. a videotape that was digitized to create digital video and audio data). File Source Mobs describe essence data, including format (i.e. WAV or AIFF for audio, RGBA, MPEG, or JPEG for video); sample/frame rate; compression type; and resolution/aspect ratio. They also provide a mechanism to locate the essence data described.
Composition Mobs describe the creative editing and composing decisions that combine individual bits of essence data into a presentation, including edits and compositing. Using metadata in the File Source Mobs, a Composition Mob can reference essence data independently of its format. In other words, applications will be able to use and display information about media data stored in formats with which they do not actually work.
Master Mobs provide an association between Composition Mobs and File Source Mobs, insulating the Composition Mobs from the detailed information about how the essence data is stored. Master Mobs can also describe audio/video synchronization, explain differences between versions of the same original essence data, and track processes such as color correction used to optimize essence acquisition or conversion.
Like metadata, essence data is also stored in objects. Essence data objects may contain a broad range of essence types and formats. These essence types include video (RGBA, YCbCr, MPEG); audio (AIFC, Broadcast WAVE); static images; MIDI; text; and "compound essence formats" (DV, MPEG transport streams, ASF). AAF also provides a general mechanism for describing essence formats and defines a plug-in mechanism that allows applications to import and export new types of essence data. In addition to the media data itself, essence objects may also include supplementary information such as frame indices for compressed data.
Beyond its Mobs and essence objects, each AAF file contains one Header object, which enables access to the file's contents. The Header object specifies the byte order used to store data in the file, the date and time that the file was last modified (or created), the AAF version number (currently 1.0), and a set of identification objects that provide information about the applications that created or modified the file. The header also contains a Dictionary object for defining classes, properties, types, data, parameters, and effects.
AAF IN DVD
Because AAF is driven by desktop video vendors, and DVD-Video is a video format, there is an understandable presumption that AAF will aid in DVD-Video title preparation. "AAF will improve the desktop video production process in general," says Microsoft DVD Evangelist Jim Taylor, "which in turn will improve tools and interoperability for desktop DVD production." However, with the exception of Sonic Solutions, the general response among vendors of DVD tools seems to be one of caution, particularly in terms of incorporating AAF into DVD-Video authoring applications. Bob Stockwell of Spruce Technologies sums up the prevailing mood when he says, "Spruce is watching AAF closely to see when it would be appropriate to implement AAF in our system."
In DVD, the interoperability facilitated by AAF would fall into two categories: between steps in the production chain--video editing to authoring, for instance--or between applications handling the same step, allowing different authoring tools to be used on the same title. Regarding the first category, some of today's higher-level DVD-Video production systems already exhibit some degree of integration between compression and authoring, in some cases passing information about audio and video streams from the encoder to the authoring application via a proprietary information file.
Even in highly integrated environments such as Sonic DVD Creator, however, the video comes to authoring already "flattened," rather than as a source video file plus a series of instructions reflecting editing decisions made in post. AAF proponents argue that the latter approach would maintain greater flexibility in authoring.
"AAF will streamline DVD publishing by transferring information vital to the DVD production process from the video editor to the DVD authoring and publishing system," says Mark Ely, director of product marketing at Sonic Solutions. Sonic already has its own AuthorScript language for passing information between steps in the DVD chain, allowing users of Avid, Media 100, and QuickTime-based video editing systems to output to DVD-Video via Sonic DVD Fusion software. But Ely says the company will make AuthorScript AAF-compliant because AAF goes further than AuthorScript alone.
"AAF is a superset of AuthorScript," Ely says. "The key additional information that AAF will pass will be editing information beyond the straight DVD features like chapter points and button links. With AAF, decisions can be made in the video editing process that can directly affect how a title is authored. Information such as pan/scan vectors, multi-angle video assembly, and multi-story/director's cut edit lists can be passed directly from the editor to the authoring tool."
Beneficial as this information interchange might be, Ian Locke, director of software business development at Minerva Systems, does not see a strong need to wait for a new format to accomplish it. "We are already working on the transfer of information from media tools to the authoring application," he says. But Locke says Minerva is "picking our battles, spending our time on the integration of existing media tools with our Impression authoring tool. For instance, we decided that the cleverest way to make DVD menus was to directly import multi-layer Photoshop files. So we implemented support for that in Impression, with all the transparency and mask information. And now we are looking at integrating things like Flash animations."
As for video, Locke says Impression is already well-integrated with video hardware from Pinnacle Systems. The companies have cooperated on a bundle including Pinnacle's DVD-1000 board, Impression, and Adobe's Photoshop LE and Premiere 5.1 RT. "All this software works together for making DVDs," he says. "The bundle also includes an export plug-in that we developed for Premiere. When you choose 'export to DVD,' you have the option of directly opening Impression and automatically creating chapter points out of the timeline markers from your Premiere project."
PORTABILITY AND OUTPUT
To Locke, facilitating metadata transfer between steps "would be a good thing, and if that's what AAF is really about we would adopt it. But what I hear from everyone who is talking about AAF is that it will allow a DVD publisher to take a project that they have worked on in one authoring program and move it to another program. For instance, you could do most of the authoring work on a low-cost Minerva Impression system, and then, because Impression doesn't yet do multi-angle, take your project down the road to a guy who's got Daikin Scenarist Professional to finish it up. There's nothing wrong with that, but for our business we have to decide whether we would rather put our engineering resources into enabling that or into developing Impression to the point where none of our users would want to move their projects."
With DVD authoring essentially at the end of the production chain, it's perhaps not surprising that lateral or upstream portability is a low priority. Even at Sonic, Ely says, the planned implementation of AAF into DVD Fusion and DVD Creator during the first half of -- will be "import to start, export only if it makes sense, which is undecided at this point. Going back into editing from DVD doesn't initially make sense."
In a similar vein, Mike Evangelist, director of U.S. operations for Astarte, adds that AAF "won't affect the output of the DVD creation process, because that realm is already standardized on the use of .IFO and .VOB files, MPEG streams, AC-3 audio streams, and so forth." However, Evangelist says that Astarte engineers believe "AAF will have some speed advantages over existing formats, which would allow a supporting application to actually get its data faster. From our perspective, this could yield performance increases in MPEG and Dolby Digital encoding systems."
With that in mind, Astarte's priorities for AAF lean more toward its desktop encoding tools than its DVDirector authoring line. "Our MPEG encoding application (M.Pack) and our Dolby Digital AC-3 encoder (A.Pack) are likely candidates since they need to be able to accept data from a variety of other applications," Evangelist says. "In fact, we will be adding AAF support to M.Pack as soon as possible--in the next two to three months."
Despite this acknowledgement of AAF's utility, Evangelist does not expect the new format to usher in the age of seamless interoperability anticipated by its proponents. "Even if AAF gains wide acceptance," he says, "it will not mitigate the need for content-creation tools to accept all the existing formats as well. The addition of the AAF format simply means that DVD producers can feed in an even wider range of source material to the process. Such a standard expands the user's choices, but it doesn't necessarily simplify anything."
Companies Mentioned in This Article
1 Park West, Tewksbury, MA 01876; 800/949-2843, 978/640-6789; email: email@example.com; http://www.avid.com
364 Wildwood Avenue, Birchwood, MN 55110; 651/653-6247; Fax 651/653-6495; email: firstname.lastname@example.org; http://www.astarte.de/dvd
Minerva Systems, Inc.
1585 Charleston Road, Mo
untain View, CA 94043; 800/806-9594, 650/940-1383; Fax 650/940-1450; email: email@example.com; http://www.minervasys.com
101 Rowland Way, Novato, CA 94945; 415/893-8000; Fax 415/893-8008; email: firstname.lastname@example.org; http://www.sonic.com
Philip De Lancie (email@example.com) is a freelance writer covering media production and distribution technology based in Berkeley, California.
Comments? Email us at firstname.lastname@example.org.