MFF2019(C)STUDIONUNU-189 (1).jpg
VRT Sandbox, start-up & scale-up accelerationAccessibilityProduction workflow & processesData, AI & ML

Limecraft and automatic subtitling

May 01, 2022 - September 01, 2023

Optimising the Natural Language Processing to create perfectly segmented and timed subtitles 

Project Inquiries

Mathias Teuwen


The challenge: maximum accessibility of content in a digital-first strategy

In the management agreement, the subtitling service aims to subtitle at least 90% of the content. However, with VRT's digital-first strategy leading to an increase in various content types such as short-form social media content, podcasts, vodcasts, and channels starting broadcasts earlier, there arises the question of how the subtitling service can handle more content without increasing the workload.

How can VRT subtitle 90% of its content in a digital-first strategy without increasing the workload of the subtitle department?
Walter Couvreur, Head of Media Access Services VRT

The solution

Through the European accelerator programme STADIEM, we discovered a solution for this issue with the Belgian scale-up Limecraft.

Limecraft addresses the challenge by providing online collaborative workspaces tailored to the creative industry. This enables them to efficiently meet the increasing demand for subtitling in Dutch and other languages. Limecraft enhances Natural Language Processing (NLP) to automatically divide the transcript into precisely segmented and timed subtitles, adhering to the broadcaster's specific brand guidelines. Furthermore, Limecraft seamlessly integrates transcription and subtitling tools into the existing editing environment, eliminating the need to switch between different applications.

The case

In this case, the subtitling service defined a clear scope. They are seeking a specific solution for automatic subtitling for short-form content (<5 min) created by content creators from the radio editorial teams and VRT MAX. Through automatic subtitling, they aim to subtitle a large volume of short-form videos without relying heavily on the subtitling service.

An essential requirement for the solution is its full integration into the work environment of the content creators, specifically Adobe Premiere in this case.

The test case unfolded as follows:

  1. Match phase: Limecraft pitched to VRT, and a Letter of Intent (LOI) was signed with the subtitling service.

  2. Develop phase: A scoping workshop was conducted, and Limecraft developed and demonstrated the plugin for Adobe Premiere panel.

  3. Integrate phase: Workshop and demonstration of the Limecraft plugin for content creators at VRT.

  4. Pilot phase: Content creators tested the plugin in a real-time environment to evaluate its effectiveness.

The results

50% of the respondents mentioned they manually subtitled before the test case. However, all respondents indicated that through automatic subtitling, they save an average of 60% of their time spent on subtitling. Further experience and sharing best practices in automatic subtitling could lead to increased efficiency.

After the STADIEM program, Limecraft successfully rolled out its Avid-integration at Sweden’s national public TV broadcaster SVT.