Digital Video, the Final Frontier

If you equate video with Hollywood, the richness of primary source footage that is now available in digital form will come as a surprise. The Internet Archive alone has well over 2000 titles and provides nearly unrestricted access to these films, fueled by the goal of encouraging "widespread use of moving images in new contexts by people who might not have used them before." Other providers of primary source footage include the Television Archive, American Memory at the Library of Congress, and American Museum of the Moving Image.

In libraries like ours—the Clemons Library at the University of Virginia (UVA)—that are committed to digital collection development and that have academic communities hungry for motion media resources, building a digital video collection is an obvious step. Although the costs in terms of infrastructure and support are certainly high, the benefits to our user community are becoming increasingly clear.

Appreciation for the scholarly value of film and video—as art, as cultural documentation, as a teaching tool, as a historical record—is high in the academic community. Making online digital video widely distributable and shareable has important consequences in our support of innovative teaching and research. Digital video supports new modalities for learning and gives scholars new forms for communication.

Digital video appears with increasing frequency in online scholarly journals, particularly as university presses experiment with new business models and forms of distribution. According to the University of Virginia Press Electronic Imprint, the possible inclusion of multimedia is one of the defining factors of the "digital scholarship" phenomenon. Postmodern Culture makes this explicit on its title page: "As an entirely web-based journal, Postmodern Culture can publish still images, sound, animation, and full-motion video as well as text." And the Journal for Multimedia History includes embedded video content in many of its articles.

A quick survey of available content shows a wealth of new offerings, with providers experimenting with models for distribution and use. We can expect to see these models evolve rapidly in response to commercial, legal, and technological developments.

The stream of commercially available resources will likely flood, as infrastructure costs drop and broadband access for end users becomes more commonplace. Feature film distribution has finally reached the Internet via Movielink, which allows the user to rent and view downloaded movies through its proprietary player. The documentary world is not far behind: Films for the Humanities & Sciences, a vendor for educational media, now offers digital licenses for many of its titles, which are sold as "hard-copy" (i.e., VHS, DVD, or DV-CAM) and then licensed for online distribution.

The world of "external" content is steadily becoming more interesting, varied, and accessible. But what about the world of "internal content," unique or rare footage already part of the library's collections? Many of our institutions hold rich repositories of rare footage in various forms, begging for digitization and online distribution. Some of that content provides matchless documentation of the life of the institution, e.g., notable events, lectures and performances, or samples of student and faculty life. As an example, William Faulkner's teaching tenure at UVA is documented on both film and cassette tapes. Although amateurish in terms of production values, the video clip "Faulkner at Virginia" offers the rare experience of hearing Faulkner read aloud from one of his works.

Digital video as an effective tool for both institutional and historical documentation is evident in another UVA project, Explorations in Black Leadership. This site documents a public program, cosponsored by UVA's Institute for Public History and the Darden Graduate School for Business Administration, that explores leadership issues by delving into the lives and careers of prominent African American leaders. What could be a more effective way to understand these issues than to watch and hear these individuals tell their own stories? The Robertson Media Center (RMC) assisted with the digitization and presentation of the video histories featured on this site.

Primary source footage culled from our own university archives achieves new prominence when used in innovative ways by academic faculty. One excellent example is William Thomas's project for the Virginia Center for Digital History, "Television News of the Civil Rights Era 1950–1970." This project draws heavily on the RMC's film archive of news footage from TV station WSLS in Roanoke, VA. Other projects in the pipeline at the RMC include digitizing selections from the hundreds of hours of footage that served as the basis of William Elwood's documentary film The Road to Brown: The Untold Story of Who Killed Jim Crow.

The full potential of digital video in academia may be reached by putting digital video cameras into the hands of faculty members as part of the research process. An astonishing example of this is the Tibetan and Himalayan Digital Library, an international collaboration led by UVA faculty member David Germano in partnership with the library. Germano and his team have created hundreds of hours of born-digital video documentation of Tibetan culture. According to Germano, "Teaching other cultures solely through the medium of the printed text is inadequate. The medium of digital video allows students to immerse themselves in the daily life of a foreign culture and to pick up knowledge tacitly along the way at various levels."

If why libraries should make digital video available is obvious, the "how" is not. There is no one answer to the question of how to build or support digital video collections. Institutions that wait for that technological solution will wait for a long, long time, missing opportunities to provide valuable resources to their user communities. A quick review of some of the solutions—however temporary—that we are implementing highlights the issues.

For starters, most digital video is not even close to being a viable preservation medium. Compression = data loss, and to a preservationist, data loss is unacceptable in a preservation medium. Think about it: one hour of 640 x 480 uncompressed video is almost 100 gigabytes. Even DV-compressed video, commonly stored on standard digital tape, requires about 1 GB for every four to 12 minutes. Any video file that is small enough to travel reasonably well across campus networks has already suffered significant data loss through compression. In the RMC we have decided to create access copies, not preservation copies. We do try to maintain DV-compressed versions of our footage, with the intention of using these tapes as "master surrogates," the source of subsequent derivative versions. This approach has served us well as we've faced the need to create several generations of derivative files for older projects.

Thankfully, the Association of Moving Image Archivists is dedicated to exploring the host of issues relating to the preservation of film and video resources.

Building digital video collections places significant infrastructure demands on the institution. Moving from creation to use, we can draw up a wish list of critical resources: high-end workstations and software to capture, edit, compress, and manage the files; servers and high-speed network connectivity to distribute the files; backup and storage systems (offline and/or online) to ensure sustainability and longevity of the files; and media playing/viewing hardware and software to use the files. The final item is arguably the most important one: staff time. Staff must be in place to design, implement, and troubleshoot all aspects of the process and to train and support users.

At the UVA library, we're slowly putting the pieces in place to handle all of these stages: all library computers have high-speed connections to the Internet, as well as sufficient RAM and processing speed to handle video playback. Most are equipped with the appropriate playing software. There are three streaming servers: a Quicktime server for prototyping and development, and both Quicktime and Real servers for delivery. We run a very fast network within our building (1000 base-T) and reasonably fast across the university (100 base-T). One staff member is dedicated full-time to patron support of digital video services; several other staff members, including a host of student consultants, have digital video support as part of their duties.

Video production takes place in our Digital Media Lab, a facility offering highly customized media support to our user community. We expect that full-scale digital video production will move within two years to our centralized Digital Library Production Service, a relatively new unit charged with creating and managing content for the digital library.

In the production of web-deliverable video, our range of source formats is predictably broad: VHS, SVHS, Beta SP, Hi-8, laserdisc, DVD, film, mini-DV, and DV-Cam. Our range of output formats for web deliverables is narrower: MPEG1, MPEG2, MPEG4, Quicktime, and Real. We try whenever possible to adopt standard, nonproprietary output formats, firmly believing that open standards should prevail in the digital library world. Our early reliance on Real is lessening; our Real server will be phased out within the year. We have never given more than a cursory look at the suite of Microsoft media products. Microsoft's continual undermining of open standards serves the academic community poorly, particularly the media users.

Defining metadata for digital video presents special challenges while also opening up worlds of opportunity. Metadata is needed to describe both external and internal characteristics of the resource. External metadata explains and maintains those critical relationships among different versions of the same content (e.g., among the 16mm source film, the DV-Cam "master surrogate," the DVD access copy, and the MPEG4 streaming file). Internal metadata allows for the description of the content of the resource at the desired level of granularity. Theoretically, each frame could have specialized descriptive metadata.

Understanding that MARC is simply not up to the job of maintaining this level of description, we look to the MPEG world for an open standards approach to the management of video information. MPEG-7 is a metadata standard that seeks to satisfy the requirements of a wide range of communities, including broadcasting and education. Eagerly awaited MPEG-7 annotation tools are starting to emerge from companies like Ricoh and IBM.

A significant step forward in the management of video information is being taken by the Moving Image Collections, an integrated online catalog of moving images held by a variety of organizations (libraries, museums, archives, and television broadcasting companies). Sponsored by the National Science Foundation, Association of Moving Image Archivists, and Library of Congress, the Moving Image Collections features a union catalog of film and video in various formats. This project is also developing cataloging utilities that allow for metadata interchange among MARC, Dublin Core, and MPEG-7—a positive development in the world of video information management.

In the RMC, we are currently managing video information in two ways: using MARC and delivering the information and media links through our online catalog, and using a homegrown product, the Audio and Video Archive Tool, to manage both technical and descriptive metadata. This tool, currently being rewritten into a nonproprietary format, will prepare video information content for inclusion in our new digital library system.

Rich metadata can open up new worlds of opportunity. At a certain level of granularity, metadata provides the framework for tools that allow close scrutiny of the contents of the media resource, allowing scholars an extraordinarily deep level of access. Imagine being able to search through multiple files carrying hundreds of hours of video to retrieve a list of indexed segments based on your search criteria. That index could be built from transcription analysis, autogenerated metadata (based on content analysis), or hand-coded metadata (provided by the librarian or researcher).

Virage, a major player in the video asset management world, includes such indexing, searching, and retrieval capabilities in its suite of "SmartEncode" products. UVA's Tibetan and Himalayan project has created its own Java-based video time-coding application that allows videos to be transcribed, annotated, and time-coded. The user can currently search the XML-based transcription and soon will be able to search the metadata and view corresponding video segments.

Considerable research energy is being devoted to investigations into autogenerated content analysis and media representation. Carnegie Mellon's Informedia provides a perfect example of this activity and a glimse into our digital video future. Informedia's prototype system uses speech recognition, image recognition, and natural-language preocessing to "read" its extensive holdings of digital video. The content is automatically transcribed, segmented, and indexed—making it available for searching and retrieval.

Developing a digital video collection is a lot like taking a family road trip. Both produce the same question from fretful passengers: "Aren't we there yet?" While a road trip requires constant adaptations to the state of the road, the change in the climate, and the availability of rest stops, creating a digital video collection requires similar adaptations to the changing technological landscape. Mercifully a road trip ends when you arrive at your destination. However, in building a digital video collection, destination is an illusion. To the question "Aren't we there yet?" the only answer can be, "No, and we're never going to be. Settle back and enjoy the ride."

While some libraries are enjoying the trip, most others have yet to begin the journey. Even when libraries have established digital collections, digital video remains both mysterious and intimidating. But when our collections include large repositories of digital video, both the business of the library and the nature of scholarship will change. We hope that more librarians will join us on this adventure. n