From QC to delivery, AudioTools Server puts AI on the case to catch language mismatches before your audience does.

AI can't replace you. But AI in AudioTools Server can make your workflows smarter.

Knowing what is on any media files that pass through your organization is a critical part of day-to-day operations. When it comes to audio, one aspect of this would be knowing the technical make-up of any files, for example, the number of channels, audio codec, bit rate, and sample rate. These types of property are usually relatively simple to determine and can be obtained with a simple file inspection. Another aspect would be knowing more specific information about what content is in the file. This could include properties such as the channel layout of the file, whether the audio is suitable for distribution, and the language(s) it contains. These types of information are harder to obtain and require a deeper analysis of the file, which can become a challenge if you have hundreds (if not thousands) of files passing through every day.

Let’s take language identification for example. Knowing what languages are present in a file is a critical piece of information. What if a file labelled as Spanish turned out to actually contain Italian? What if a file supposedly containing English on the first two channels and French on the next two was actually the other way around? The consequences of transmitting the wrong language don’t need to be exaggerated. 

Identifying what languages are present on a file has traditionally been a manual task, requiring a human operator to open the file and review the audio on each channel. This is both a time-consuming process and one that is open to error. It relies on the operator being able to correctly identify the languages present, which may be challenging if the operator is not a native speaker. Even if the language is correctly identified, there is still the risk that the wrong language metadata tag will be assigned. Clearly a more modern approach is required.

How AudioTools Server Automates Language Identification

The Language Detection option available for Telos Alliance’s AudioTools Server provides just this, forming part of the next generation of AI-driven processes available with the AudioTools platform. The Language Detection module uses the Fraunhofer Institute’s “LanguageID” technology, which employs the latest machine learning processing to intelligently analyze the audio and report the languages present, choosing from one of the hundreds of available language options.

The detection process breaks the audio down into 10-second segments, reporting the most likely language candidate for each segment. Once the analysis is complete, the results for the different segments are compiled and the most likely language determined. For users who require a more granular approach, a full report can be produced detailing the languages present at each point. This can be useful if you want to make sure that the language doesn’t change partway through the file, or cross-check any points where a different language is present to make sure subtitles are present.

Here’s the secret sauce…

What really sets the AudioTools Server Language Detection apart is the ability to take the result of the analysis and then use it to drive other processes. This could be as simple as checking whether the detected language matches the one expected and quarantining/flagging any files where this isn’t the case, or it could be something more advanced. 

To illustrate this, let’s imagine that you have a workflow that produces deliverables comprising multiple languages. With the AudioTools Language Detection option, incoming files can be analysed and the languages detected on each set of audio compared to the expected mappings. If the language layout is correct, the file can be passed straight through to some later stage in the workflow (such as loudness processing). If one or more of the expected languages are missing, the file can be rejected. If the right languages are detected, but arranged in the wrong order, then the workflow can automatically remap the tracks into the correct order. By combining Language Detection with other processes, such as Channel Detection or Channel Management, it is possible to build very powerful workflows that ensure you consistently deliver correctly laid-out audio with no human intervention.

Language Barrier? Not Anymore.

From now on, language differences aren't an obstacle to your content.

Like all AudioTools Server processes, Language Detection is highly scalable. So whether you have just a handful of files that need checking a day or thousands, the system can grow to meet your needs. With a module available within our award-winning WorkflowCreator workflow design tool, it is very easy to start building workflows that include language analysis as part of their processing.

To learn more about Language Detection and the other tools available with AudioTools Server, please contact your AudioTools Server representative. We’d love to hear from you.