Creating and translating audio descriptions with the help of Frazier

International audience ; Keywords: audio description, translation, shows, corpus, French, GermanThe French-German research project TADS (Translation of Audio Description Scripts) aims to examine the translatability of audio description scripts. The team expects that machine translation tools can help create more efficient workflows in order to meet the increasing demand for audio descriptions (AD) on television and online streaming services (Jankowska, Mrzyglodzka, and Chociej 2015). One of the main research questions is in what way AD scripts can be transferred into different languages and cultures, taking into consideration the norms and guidelines of each country and broadcasting station as well as culture-specific and linguistic elements (Zago 2019). For this purpose, the researchers have started to gather a comparable and a parallel corpus containing AD scripts of French and German television shows, building on results by Reviers (2018). This corpus can be accessed on TXM (Heiden, Magué, and Pincemin 2010).The German company VIDEO TO VOICE, one of the cooperation partners of the project, provides the researchers with its AD software Frazier (Pajonczek and David 2019). This software allows a fully automated generation of an audio described video, including a variety of artificial voices and automatic translation. In this paper, several functions of Frazier will be presented, amongst them the insertion mode, the annotation mode, the automated ducking function (which decreases the volume of the video’s soundtrack whenever a passage of the audio description script is heard), and the translation options. The product can be adjusted to different broadcasting standards used online and for television in the respective countries. The researchers plan to compare the finished products with data from the AD corpus, and they aspire to use the former for accompanying reception studies with the visually impaired target group(s), who are asked to evaluate the acceptability of different audio descriptions automatically .