How does work under the hood?

Wrap all metadata is a standardised container with a standardised identifier and a standardised timing model. The identifier is used to find a mapping from the contents of the container to the standardised timing model

Core principals

The core technology works like this:

  • Sender
    • Categorize metadata as (binary or text)
    • Categorize metadata as (clocked or unclocked)
    • Wrap metadata in an MXF container / packets according to category
    • Add metarex identification (mrxId) metadata & timing information
    • Check / Register the mrxId in a public register so that anyone can figure out what the metadata is
    • Map the mxf packets into a network transport for real time distribution if needed
  • Receiver
    • Unmap network transport into a sequence of mxf packets
    • Store / preserve /mux / process the mxf packets as a file or stream
    • Get mrxId registration info from the register
    • Process metadata / download a plugin / display a message based on the registration information provided by the owner of the registration record.



Looking at the tools available for packetizing a collection of abstract metadata elements, there are a few requirements that were needed:

  1. identification hierarchy aware of classes of data and instances of data
  2. clock metadata with precise clocks
  3. lock metadata timing to existing media clocks e.g. video / audio
  4. synchronize metadata with other streams e.g. video / audio / rtp / gps
  5. packetize text based metadata
  6. packetize unsynced timed data e.g. subtitles / elapsed time / script progress
  7. carry any metadata type without transcoding
  8. can securely encrypt content with an established Key Management Regime
  9. extend the system as it matures

Many candidates were considered including tar, zip, quicktime, matroska yet all of them lacked one or more of the key infrastructure elements to satisfy the requirements. MXF is 20 years old, widely deployed, has a sophisticated timing model, encapsulation model, an ID infrastructure that gives every MXF file in existence a unique number, an extensible data model, a field-tested encryption model, and has extension mechanisms including the ability to define in-file dictionaries for sophisticated parsing.

Basically, MXF already has the tools needed and the other options require significant work to create a general purpose global solution.

The MetaRex MXF profile

Some features of the MXF profile are extensions of the MXF-live profile that was demonstrated in 2019 at the ARRI Broadcast Day .

The MXF profile document is a work in progress - if you want to contribute to it or have issues with the design approach then please [contact] us or join the project to help us complete the work an give away the software.

Some desires of the final MetaRex MXF profile:

  • it must be possible to serialize & stream an mrx.mxf over a network
  • it must be possible to concatenate a stream of mrx.mfx packets to an atomic OP1a MXF file
  • every partition of an mrx.mxf file must only contain one thing(Header, Index, Body, GSP, Footer etc.)
  • an mrx.mxf file must contain a minimum amount of descriptive metadata (DM-MRX) to enable rich identification via the MetaRex register.
  • an mrx.mxf file must contain a minimum amount of descriptive metadata (DM-MRX) to enable an appropriate time stamp for the metadata contained e.g. NTP derived time of day in ISO 8601 Timezone format.
  • an mrx.mxf file might contain richer MXF-native metadata.
  • an mrx.mxf file might contain richer timing metadata.
  • an mrx.mxf file might contain richer meta-metadata using KXS ( SMPTE ST 377-2 ).

MetaRex-MXF (starting doc for WG)

A version of the MXF-Live specification to start the Metarex project. This version will be updated and edited as the project proceeds.

Last modified January 31, 2023: blog dates fixed. MXF book links (ceb1cf9)