Introducing Computer Assisted Localisation

Maarten Verwaest
September 9, 2018

More and more content is being distributed globally and requires localisation. But because the cost of localisation is a serious burden on the budget and because it can take a lot of time, we took the challenge to radically automate the process.

Yet we have to manage expectations. Radical automation is not the objective, because machines make interpretation and translation errors. However, statistics show that we can reduce the cost by 50% to 80% depending on the quality of the audio, and as such improving the chances for distribution of valuable content that would otherwise remain undisclosed.
Localisation services are currently available in private beta. If you have an interest to join the test pannel, please drop us a line via info@limecraft.com.

Why automatic Subtitling

Producers of audiovisual content, including Film, Television and Corporate Video, not only consider subtitling as an essential part of the production process, it has become mandatory in many places to provide subtitles for reasons of non-discrimination. To make things even more complicated, content is being distributed globally so it needs localisation whereby the audio is translated in subtitles of other languages.
However, the cost of manual subtitling and localisation is significant. Depending on the type of content, it takes 10 to 20 hours hours of work to manually edit subtitles of a 50′ documentary or drama. So at Limecraft we asked ourselves the question what it took to radically automate the process.

Adding Machine Translation

To automate the subtitling process, a number of different challenges faced our development teams. Turning audio into text is the easy part, as you can buy pretty good automatic transcription services from Limecraft or other parties. Turning the script or the transcription into properly looking subtitles was the more challenging part, as we explained in an earlier blogpost.
So now we added machine translation to the pipeline. With a single click, you can instruct us to translate the transcription from the original language to the destination language.
Next, because we maintain the timing during translation, you can cut the translated transcript into localised subtitles. All are displayed side by side with the media, so you can play the sequence with subtitles, check the quality and efficiently correct where necessary. When finished, you can export the subtitles in any format of choice (SRT, STL, EBU-TT-D).

Key Benefits

Statistics show that you can realise cost savings of 50% to 80%, depending on the quality of the audio and the grammatical quality of the spoken language. To get the most out of it, we advise to use a proper microphone setup, to avoid cross-talk during interviews and to give instructions to speakers to use well-formed sentences as much as possible.
Also, from the feedback we got to date, we understand that the shortest possible cycle time is at least as important as cost reduction. We trust that journalists and reporters can now produce subtitles themselves, as we took great care of the usability, and as such make their story available for distribution minutes after finishing the edit.