While interpretation into one or two languages has typically been provided at EBU Assemblies, this tool promises to enhance the experience of every delegate, with anglophones too potentially appreciating having a live transcription available. The transcriptions and translations will be made available through a companion web-based interface that delegates can view on their digital devices.
EuroVOX has up to now been better known for applications involving file-based content but was also designed to handle live content. The team created a workflow to enable the service that, thanks to the EuroVOX API, can take advantage of different technology providers for both transcribing and translating. This capability means that services are not limited to the languages offered by a single provider, and new providers can be integrated without any additional work for the implementers, for example if the quality improves or the pricepoint decreases for a given prodvider.
This flexibility has been demonstrated with the recent inclusion of OpenAI’s Whisper to the set of available providers in the EuroVOX API for file-based content. With Whisper using the same Large Language Model (LLM) as ChatGPT, early tests show that it brings significant advantages for many broadcasters using EuroVOX: better quality transcription results, handling of mixed-language content, and speed and price improvements.
If and when Whisper is available for transcribing live content, its integration into the EuroVOX API will mean it too can be applied to the live captioning workflow without any additional changes.
While the immediate aim of the recent work is to provide a useful service to Technical Assembly delegates, the event also provides an opportunity to demonstrate how realtime transcriptions can be useful for both original language captions and also translations. The new functionalitiy will be fully integrated into what will soon be rebranded as EuroVOX Studio. This will allow EBU Members to translate and transform both file-based and live content in a straightforward and coherent way, leveraging AI to assist in transcriptions, translations and revoicing.