As a continuation from the previous post, which lays out the basic reasoning for the metadata conundrum, we’re going to look at the the multiple points at which metadata is lost.
The first phase of metadata loss - acquisition - is one that has had the most progress in the time from which this information was first presented (1999).
Back in 1996, I began the quest for the perfect metadata system - one that would track stock or syndicated clips throughout the production and distribution process. By 1999, during consulting work with a client with a compelling need to track metadata across the total production landscape, we prototyped a metadata tracking system that MCI then turned into a media asset management (MAM) service.
The core prototype had a part that was never built into the MCI system - a measurement tool designed to differentiate between stock footage downloaded for introduction into a rough timeline versus content that ends up in the final product.
Consider this scenario that was first proposed in a 1999 white paper.
An independent filmmaker, who is creating short-form content for the web, decides she needs a montage of video clips that show the history of rocket launches. She finds stock or historic content that includes captured German V-1 and V-2 test footage, Goddard’s initial rocket tests, Apollo and Gemini lift-off sequences and several clips of recent shuttle launches and landings.
She likes the footage enough to pull down the “trial” versions (low-resolution and watermarked, of course) but then finds that these versions have key resolution deficiencies and watermarks in places that may limit her FX compositing. As such, to work around both limitations to the creative process, she’s compelled to download (and pay for) higher-resolution files.
After sequencing the FX compositing, she makes a creative decision to pare back the number of clips used. This paring could be done for a variety of reasons: she sees the original resolution on several of the clips isn’t going to allow complex compositing; she finds the sequence needs to be shorter; or she sees the shuttle she’s chosen is one that has been de-commissioned. Regardless of the reason she
At this point, though, she’s already paid for the clips and can’t exactly ask for her money back. The requirement to pre-buy stock content, regardless of whether it’s used or not, severely impacts other portions of the limited indie production, overwhelming an already tenuous production budget.
Sound farfetched? It happens every day in a variety of productions, as old business models meet new business realities.
The old business model of pre-paying for stock footage holds on for good reason: the companies selling stock or historic content can’t accurately track where their content is being used, either in finished timelines / productions or in types of distribution. That’s the essence of the work being done to maintain metadata throughout the production and delivery process.
Why not go with an even older business model - the “honor system” that was part of music production libraries when I was in film school, where music usage was reported by the content creator after the fact? For those of us who used these tools, there was a double penalty: pay for the music library up front and then pay for usage, with uncertain results at best for libraries like DeWolfe as independent content creators didn’t have much control over where their content was shown. The rise of royalty-free music libraries (and the inherent deficiencies of putting the burden of usage tracking ad infinitum on the content creator) deep-sixed the honor system as the proliferation of new venues and distribution models blossomed at the dawn of the Internet era.
Still, the basic problem remains: clips dropped into a non-linear editor’s bin lose about 60% of about of the pertinent metadata; clips then dropped into the timeline lose an additional 20% of metadata; and clips output to a rendered file (or tape) lose 99.999% of metadata about its sub-components.
The use of a manufacturing term to describe clips on the timeline is intended to make this point: video and audio editing / distribution - almost exclusively performed on computers and digital media - has less metadata tracking capability for its subcomponents than the average 2000 model year car rolling off a Detroit assembly line. Even the glass on the car can be traced back to a particular vehicle, assembly line and date of creation. Sheesh!
Additional information regarding acquisition, production and distribution metadata, including additional excerpts from past white papers and reports (as well as the foundation for a new set of white papers) will be posted over the next few weeks.
Posts
All Comments
