Thursday, 23 November 2017

Metadata workflows: One format to rule them all

IMF is coming to broadcast. What does this mean for post workflows and will it achieve the simplicity it intends.
By April next year there will be a new file format to juggle with. This one is intended to be the one format to rule them all. SMPTE is working with the Digital Production Partnership (DPP) on joint development of an Interoperable Master Format (IMF) specification for broadcast and online.
The original SMPTE standard, ST 2067, developed in 2012, dealt with file-based interchange of finished multi-version audio-visual works. That dealt with multi-language requirements, including subtitles and closed captions all of which can be handled within one large IMP – the interoperable master package.  
The DPP then took the lead in drawing up an IMF for broadcast, in co-operation with the North American Broadcasters Association (NABA) and the EBU.
For broadcasters there are two primary use cases: incoming, meaning buying content masters for further compliance processing; and outgoing, which is sales mastering. The goal is to implement a system that addresses the myriad metadata requirements of television and OTT while fitting into broadcasters’ sizable existing archives of content.
However, the jury seems out on how successful the original IMF has been. Paul Mardling, VP of Strategy, Piksel says “extremely successful” but others are less sure.
“The original vision for IMF was that movie studios would use it as the primary format for archiving content and delivering it to the supply chain,” says Dominic Jackson, Product Manager, Enterprise Products at Telestream. “So far there may be some archiving going on, but the use of IMF as a ‘live’ delivery format isn’t really here yet. We are mostly at the point where studios are distributing ‘test’ packages to ensure that recipients can handle them correctly.”
Dalet’s Chief Media Scientist, Bruce Devlin also claims take up within the Hollywood community, However, there is still a missing unifying aspect to IMF. He cites two major components behind this inability to truly make IMF a mass-market format.
“Each studio still has their own IMF delivery standards, both input and output specifications,” he says. “The broader content creative and delivery community feels that IMF is really more like 5-6 different flavours of a similar standard, since they have to make IMF Flavour A for Netflix, IMF Flavour B for a major studio and IMF Flavour C to feed their finishing/transcoding tool.”
The second missing he feels is the lack of tools able to handle IMF. “Componentized media workflows like IMF are very powerful and drastically simplify operations, but they are very complex in the back end. It requires a good platform and adapted management tools to enable simple, cost-effective solutions to ingest, manage, search/find/retrieve and transform IMF for the necessary workflows. And very few platforms have developed the data model and toolset.”
Such tools are expected to move out of testing into actual use cases and production over the first half of 2018. As they do, there shouldn’t be too much difference between the core workflows for IMF version 1 and the broadcast/online version but there are changes.
The principle differences is the need to transport ad break information to support stitching of assets from outside of the IMF container at playout.
“This opens the possibility of creative teams defining different break patterns and cut points in the asset depending on the required ad payload and playout length requirement,” explains Mardling. “There’s also the need to include the relationships between assets in different IMF containers to allow series and episode information to be transmitted within the container.
“There are effectively two approaches to IMF in the workflow – early or late stitching,” he says. “In an early stitching scenario the broadcaster flattens the IMF file to the required formats on receipt. This allows the rest of the broadcast workflow to remain more or less unchanged. In a late stitching scenario the IMF is processed through the workflow and in effect any edits are simply reflected in an updated CPL. Flattening does not need to occur until final playout, potentially on the fly.”
Any new version of IMF needs to be backwards compatible with current implementations. However, the majority of broadcast archives are currently flattened files, often in multiple difficult to track versions. According to Mardling, the principle issue will be in the requirement for updating tooling to work with IMF and the risk that rendering technologies may become obsolete.
“I’d expect groups like DPP to work on defining standard metadata sets for specific markets or groups or broadcasters,” says Jackson. “These metadata sets will also need to be extensible to include specific metadata requirements for individual broadcasters.”
If IMF for broadcast is going to work, broadcasters and the kit suppliers will be required to introduce new ways of handling files.
“Most of the existing tools that broadcasters use are not necessarily very IMF friendly,” says Jackson. “The asset management systems, transcoders, edit systems and the like are all likely going to need some upgrade to deal with IMF successfully. Just the implications of the fact that IMF is a group of files rather than a single one are going to be significant for broadcast supply chain, and that’s just the start of it.”
Codec issues
Deciding on a primary codec for the format seems too thorny an issue to be sorted. “ProRes in IMF seems to be here to stay but I doubt that will meet everyone’s needs,” says Jackson. “Codec choice is too challenging to agree on.”
Mardling believes the issue is a storm in a teacup; “As with other similar standards, shims will be required to ensure interoperability but this should not be a major impediment to adoption.”
The codec issue arises when broadcasters need to know that IMF will be backwards compatible with their archive.
“Given that IMF itself is a standardized format, and self-referential toward the essence storage, the main concern broadcasters should have with their archive is codec choice,” says Devlin. “Will my chosen codec be usable 10-30 years in the future? You would be hard pressed to agree that any codec designed for archive, beyond JPEG-2000, would answer that question as a yes.
“So, you have a trade off that occurs in broadcast archives that is different from studio archives. Whereas studio archives are generally more concerned with long term preservation, repackaging and resale, broadcast archives are generally more concerned with enabling fast turnaround workflows, with long term preservation a secondary concern.”
Additionally, for archival, IMF provides a mechanism for embedding or linking private metadata into compositions. This allows ‘helper’ metadata to be included in the archiving process for applications such as re-creating MAM records in business continuity applications. Devlin says IMF looks very attractive for archive applications and there is a special application App#4 for cinema archiving. That is currently being used as a template for gathering TV archive requirements.
One of the chief benefits of adopting IMF for broadcast/online is the promise of more automated (efficient) workflows. Mardling cites the example of a current affairs programme containing content that needs to be cut/edited at the last minute for legal or compliance reasons. Rather than having to return to the edit suite and multiple new versions be produced, a simple change to a CPL can update multiple versions ready for playout.
Jackson remains unconvinced. “The one area where it brings obvious efficiency is for a broadcaster who has a need to archive multiple versions of a piece of content,” he suggests.
His suspicion is that many, possibly most, broadcasters will not adopt broad usage of IMF. “There are a small but vocal group of proponents but I suspect ultimately that many will see IMF as introducing more issues than it solves.”

Related IMF (version 1) technologies
Ownzones Media Network has released Ownzones Connect, a platform which ingests, categorizes and stores multiple media files in one location stored in the cloud. It also includes a dynamic creation of video jobs based on templates, and the introduction of smart agents for proactive metadata optimisation.
Version 5.9, of Rohde & Schwarz’ CLIPSTER offer a complete workflow from mastering, versioning to merging and refining IMF packages. CLIPSTER can be used arrange the various video and audio tracks, to create the master package, to generate any number of versions and to merge or to split the packages.
Prime Focus Technologies supports IMF within CLEAR, including an IMF Player that provides the ability to preview, playback, review and distribute over a streaming proxy. This enables collaboration and decision making in the workflow using proxies without having to necessarily access the original IMF package in high-res each time a CPL has to be played back.
Telestream support for IMF is in media processing product Vantage. It can ingest an IMF CPL as a master source input to create all appropriate outputs, and can create single segment IMF Master Packages as an output. 
Today, IMF packages are becoming simpler to create and interchange, but the lack of a fixed file or folder naming scheme is making management of IMF - especially with tens and hundreds of supplemental packages per title - more complex.
“Implementing a MAM and Orchestration platform such as Dalet Galaxy will allow you to scale and industrialize your operation and break the many IMF versioning workflows with greater ease and accuracy, and at a much lower cost of production,” suggests Devlin.
A testament to the veracity and robustness of the solution is recent Netflix compliance approval awarded to the Dalet AmberFin transcoding platform. This means that producers and facilities needing to create IMF packages for their Netflix targeted content can use Dalet IMF technology to get the job done.
Sidebar: IMF recap
Built on the proven success of formats such as DCP (Digital Cinema Packages), IMF is a media packaging format that streamlines the assembly of multiple versions of a title for any downstream distribution. Instead of a single master file – like a QuickTime or XDcam file, all IMF audio, video and text components are contained within the file as individual assets. IMF components are individually referenced by an editorial Composition Playlist (CPL). As an analogy, the assets are the ingredients and the CPL is the recipe.
An IMF can contain unlimited CPLs, each representing a unique combination of the files contained in the package – or cuts of a programme.
Instead of hundreds of separate versions of a master, users can create one IMP (Interoperable Master Package) with all the various media elements (soundtracks, subtitles, graphics, technical metadata etc.) and endless variations described by the CPLs.
The IMF format also includes an Output Playlist (OPL). The OPL is the IMF package’s technical instructions and it can incorporate additional modifications like sizing, audio, channel mixing and transcoding profiles. Additional OPL functionality for broadcast is still in development, with a range of companies, including Prime Focus Technologies, working to create the specification and tools.
Essential workflow steps for IMF for broadcast
First, broadcasters (and their suppliers) have to start with re-architecting your media handling as essential components of video, audio, captions/subtitles, metadata, rather than as an interleaved asset.
“Componentized media enables all the nice versioning and space saving capabilities of IMF, but does require that you have the ability to match up the components during your WIP phase to associate them back to the project at hand,” explains Devlin.
“In an interleaved media workflow, as has been in the norm in broadcast since the dawn of the file-based age, your main challenge is just passing a single file around between disparate processes, users and possibly, external vendors. This can be accomplished via watch folders, file acceleration services and even, sneakernet of USB keys (as in, ‘Joe, go run this USB key over to edit bay 2’), generally, without the need of a higher level management system, since the single file can be fairly self descriptive and tracked through a linear process.”
In a componentized media workflow, Devlin says, broadcasters have the same challenge of connecting processes, but also the need for a management system that can manage multiple files in a project, assign those files out to parallel media stages, and reassemble them for IMF packaging and downstream transformations.
“Finally, IMF workflows will nearly always be implemented using some sort of orchestration engine. To create an IMF composition, you need to obtain metadata from IMF Track Files, metadata from business systems and metadata from MAMs and put them all in a CPL. This aggregation of metadata from multiple sources needs to be flexible and be subject to automatic QC. This can be done at scale using an orchestration system, but would be painful if performed manually.”

No comments:

Post a Comment