More then words. Voice in audio description

Audio description research has been expanding in the last years, with different theoretical, descriptive, technological and experimental approaches. Very often the focus has been on how certain cinematographic elements can be better audio described, often by analyzing textual elements and by investigating user reception. However, there is one area in which research is still scarce: voice in audio description. Investigations in this field have mainly focused on text-to-speech audio description as compared to human-voiced audio descriptions. Guidelines often indicate that audio descriptions should remain neutral (without defining what "neutral" means), with some promoting more 'expressive' intonations for children. Still, a thorough analysis of prosodic features in audio description is still not available. Within this background, an expert in accessibility and audiovisual translation and two phoneticians have teamed up to carry out a three-step research: firstly, a literature review on what guidelines indicate in terms of prosody. Secondly, an analysis of existing audio descriptions in terms of prosodic features and, finally, a user reception study. In this presentation we will focus on the second and third step. The VIW corpus (http://pagines.uab.cat/viw/content/corpus) has proven the perfect material for the second part of our research. It contains 10 professional audio descriptions in Spanish of the same clip, the short film 'What happens while'. Professionals were given the same instructions and interestingly enough the results differ in various aspects depending on the prosodic features. Using the 10 professional audio descriptions available in Spanish, we have carried a prosodic analysis in which we have analysed speech rate and amplitude values. Concerning the third stage, an online questionnaire on user preferences has been administered online to a sample of users, both sighted and visually impaired in order to assess the naturalness and the voice quality in two different user profiles. In Language and the Media we will present the results of our research, hoping to trigger a lively discussion not only among researchers but also among professionals delivering audio description. This proposal is part of the NEA project, funded by the Spanish Ministerio de Economía y Competitividad under reference code FFI2015-64038-P, MINECO/FEDER, UE