Automation – The Next Frontier in Multimedia Localization?
Click here to close
Click here to close
Subscribe here

Automation – The Next Frontier in Multimedia Localization?

Automation – The Next Frontier in Multimedia Localization?

Multimedia LocalizationIt is an established fact that when it comes to readiness for localization, most multimedia files produced today leave much to be desired for.

Examples? There are plenty. It’s often difficult or time-consuming to identify localizable content (text). When identified, there may be no efficient way of extracting localizable assets. And once localized, re-integration of the translated assets back into the source multimedia file may be, well, painful. In fact, preparing multimedia for efficient localization may sometimes feel a bit like trying to put toothpaste back into the tube. Possible, but… Why is it so and what can be done about that?

Don’t Blame the Messenger

This is partly natural. “Single-media” such as text will normally lead to less complexity than “multi-media”. But more importantly, subsequent multilingual localization would not normally feature high in the priority list when multimedia materials are being created.

Creativity may take precedence, and the actual teams behind multimedia creation may lack the hard-won experience of technical authors and technical publications specialists (and, I don’t blame them). In some ways, the average localizability of multimedia files today may be where the more traditional formats such as software, documentation or web content were a decade ago. Granted, I don’t have any hard numbers to support this statement, but this is just the way it feels.

Enter Automation

In our experience, intelligent automation can go a long way to compensating for the way multimedia files tend to be handed off to localization. Let’s take the example of a standard multimedia file created in Adobe After Effects, the very popular tool for creating motion graphics and visual effects. The project file may be complex and contain sequences and sub-sequences with thousands of objects, including text layers or text assets.

Automation is key to overcoming such a chaos in creative environments. The typical stages where automation has a role to play include:

  • Automated asset export
  • Pseudo-localization
  • Automated asset import

There are a number of tools/workflows available that will export and/or import texts from After Effects projects. But true efficiency will be obtained by using a customized script that will produce a standardized XML file ready for a CAT tool such as Trados, rather than a huge (and messy) XML file. Such a script should also extract only unique items, no duplicates.

This script may also greatly facilitate actual identification of all localizable assets via pseudo-localization. As a rule, you will not know straight away which localizable items will actually appear in the rendered video, nor will you know the volume of text included in the project. Will everything be localized?

Running the extraction script, pseudo-localizing the XML file and reimporting the pseudo-localized file back into the project, will help out. This way you can easily identify localizable text assets, but also additional items or objects that may need localization, such as images or embedded videos. This use of pseudo-localization at this stage complements its other, typical objectives (localizability, text-expansion or character issues, etc.).

From Person-Days to Minutes

Here is a quick summary of how automation may help:

  • You have a huge project with thousands of assets: Automate extraction, remove duplicates. A manual task taking 2-3 person-days may be reduced to 1 minute needed for extraction.
  • Extent of localization is unknown: Pseudo-localize, render the multimedia file and review the pseudo-localized version.
  • Preparation for translation in CAT is cumbersome: Extract standardized data into an XML file and use standard (!) parsing.
  • Integration of localized content is time-intensive: Ensure localized assets reimport automatically. A manual task taking 2-3 person-days may again be reduced to 1 minute needed for importing. Verification and bug-fixing iterations may then be reduced to one final review.

In our experience, the bottlenecks to efficient localization lie in manual tasks and verification tasks performed late in the process. Remember, what is prepared manually has to be re-integrated manually and verification may take more than a few rounds.

Interested to learn more? Automation in multimedia localization was the theme of Moravia’s webinar Clearing Roadblocks to Efficiency, which dealt with some of the frequent multimedia localization issues which tend to make multimedia localization less efficient than otherwise possible.