Project's focus was to replace relatively complex manual routine production tasks through quality-assured automated processes. Periodical tasks which had to be processed by the production team manually at DigAS and Sequoia audio editing suites should be replaced by a fully automated software solution.

To make even more complicated, the system has to be able to work correctly on very challenging material, like classical music. As an example, with classical music it has to automatically detect whether the length of an applause is ready for broadcast or has to be cut, or if pauses in sentences had to be shortened for material to be broadcast. The audio material has to be cut with a precision of only a few milliseconds to the modulation start point, to enable playout without sounding as if it was cut in the first place. Some of the requirements for the software were that it is able to plan and do needed cuts on its own, and that fades cannot be heard. Additionally, corrupt files have to be detected automatically, for example if there is an unexpected modulation break.

An essential requirement was to process lossy encoded MPEG Layer-2 formats directly (without decompression to the linear PCM), to eliminate quality loss through re-encoding. For this, Cube-Tec used a mechanism which is able to do cuts and fades directly in MP2 files. Of course, overlength multichannel MBWF and RF64 files had to be processed correctly as well.

The system has to classify its own work quality, so that - if any file falls below a specified confidence interval - it is able to transfer that file (together with all analysis already done) to a human worker for inspection. After this human task the material can be added to the automatic workflow again easily. Through this seamless human-machine-integration it is possible to only have manual work where it is unavoidable and provides real added value, while every automatable process is actively unburdening human workers.

What would broadcast finalization be without loudness measurement and correction? Cube-Tec already introduced on IBC 2010 its solutions for file based loudness correction following the EBU R 128 standard and was able to use proven software modules from its CubeWorkflow and DOBBIN Media Rendering Farm solutions for implementation of this project.

There were less than six months between planning and approval of this project, which was possible only through precise planning of details by complex process modelling and careful creation of target specifications.

Many simple automation tasks are already implemented by media companys. Now things get exciting, as future automation projects have to be able to transfer much more demanding tasks and more varied manual tasks to machines. The same importance is needed for seamless embedding of human workers into the process chain for special task parts, while supporting them in the best ways possible. Here we have a clear advantage through our agile CubeWorkflow system. I really appreciate this project and the trust in our team. Only by the good support of the project members at Bayerischer Rundfunk, the project could be supplied on target and on time.
- Jörg Houpert, Head of Technology, Cube-Tec International GmbH