At some point, independently of what kind of media supply chain you are working in, media and metadata will arrive in some form. Some examples.
- A production team will need to log incoming raw material from the daily shots.
- A VOD delivery platform will acquire new titles by file transfer of video, audio, and associated metadata by an online delivery service – or only by SFTP.
- A live streaming service will need to store clips and metadata from the live broadcast feed.
The process of acquiring and ingesting media can be challenging in many ways, but is arguably the most important step in the workflow chain. Technical or editorial faults or errors in metadata or the media itself either introduced or allowed to pass through at this stage will cause greater problems further down the chain.
Your media supply chain needs to be able to decode and work with your media files to start with. This, most of the time requires a first technical analysis and potentially a QC check to determine what transformations need to be applied to the media files to meet the “house” standards. This may include transcoding, using a service like VidiCoder, to an in-house, or mezzanine format, or other media transformations. In some workflows, a low res “proxy” version of the media also needs to be produced to enable desktop or remote viewing and editing.
The same applies to the metadata information that many times needs to be transformed to conform to in-house metadata schema.
A production workflow will add metadata to incoming media in a logging process. Traditionally the logging process in an ingest environment has always been a largely manual task, using tools such as MediaIngest and MediaLogger, adding descriptive metadata to the media as a whole, but in many cases also identifying “temporal metadata” describing actions or objects that appear for a specific duration in the media.
However, with machine learning, this is about to change dramatically. Today, a machine learning algorithm, such as the cognitive services available in the VidiNet, can detect objects, people and spoken language and make that available as searchable metadata. While not entirely replacing the manual process, such services can significantly reduce manual effort while signficantly extending the metadata set.
Ingest workflows at the distribution end of the media supply chain have similar requirements, although the relevant metadata and media formatas will likely differ. Analyzing the media and making sure that media and metadata conforms is still a critical step, but with different criteria, such as regulatory compliance. Automated analysis and cognitive services can again be applied here. Enabling new and essential processes like automatic trailer editing, subtitling, branding detection, scene detection for AD insertion, and much more.