Audio describers of feature films are writers–or let them call storytellers–specialising in audio-visual translation targeted at blind or visually impaired people. One of their professional skills is the ability to imagine the information needed by this sort of audience: one thing is to help them catch the story told by a movie; another is to make them feel touched when listening to its audio described version. The second requirement can be pursued through audio describing mental states like fear, joy, despair, or hope, most of them left open to interpretation, thus risking not to be objective when mentioning them. As a matter of fact, mental states (Oosterwijk et al., 2012) can be classified as episodic film events which may be audio described or not, depending on available space and on other criteria like the macro-function of the film genre (Vercauteren, 2016; Mazur, 2020). Our presentation does not include audio descriptions of stable emotional features of film characters, like "a cheerful soul" or "a cry-baby", less subject to decision-making. Within the context of the TADS project (Translation of Audio Description Scripts), we discovered differences regarding the treatment of mental states between the audio described German feature film In den Gängen (Stuber, 2018) and its version dubbed in French, Une valse dans les allées. These differences concern the choice of scenes displaying mental states and the linguistic means used to audio describe them. We first will present a way to identify mental states in the audio description script of the movie In den Gängen and the one of the version dubbed in French. We will show how we used the corpus tool TXM (Heiden, 2010) to search the data for linguistic forms (e.g. adverbs or adverbial adjectives) or patterns (e.g. attributive expressions) representing feelings. Ott's psychological categories, which list emotions and their physical manifestations (Ott, no date), are used as a starting point to classify and analyse the occurrences obtained. In a second step, with the help of the transcription tool EXMARaLDA for multimedia data (Schmidt and Wörner, 2013), we will compare the means used in our German and French audio description scripts to refer to mental states. References Heiden, S. (2010) ‘The TXM Platform: Building Open-Source Textual Analysis Software Compatible with the TEI Encoding Scheme’, in Proceedings of the 24th Pacific Asia Conference on Language, Information and Computation. PACLIC 2010, Tohoku University, Sendai, Japan: Institute of Digital Enhancement of Cognitive Processing, Waseda University, pp. 389–398. Available at: https://www.aclweb.org/anthology/Y10-1044 (Accessed: 20 November 2019). Mazur, I. (2020) ‘A Functional Approach to Audio Description’, Journal of Audiovisual Translation, 3(2). doi:10.47476/jat.v3i2.2020.139. Oosterwijk, S. et al. (2012) ‘States of mind: Emotions, body feelings, and thoughts share distributed neural networks’, NeuroImage, 62(3), pp. 2110–2128. doi:10.1016/j.neuroimage.2012.05.079. Ott, H. (no date) Reconnaitre les émotions. Available at: http://www.ieccc.org/IMG/pdf/Reconnaitre_les_emotions_peda.pdf (Accessed: 19 June 2021). Schmidt, T. and Wörner, K. (2013) EXMARaLDA. Oxford: Oxford University Press. doi:10.1093/oxfordhb/9780199571932.013.030. Stuber, T. (2018) In den Gängen. Sommerhaus Filmproduktion, Mitteldeutscher Rundfunk (MDR), ARTE. Vercauteren, G. (2016) A Narratological Approach to Content Selection in Audio Description. Universiteit Antwerpen. Available at: https://www.semanticscholar.org/paper/A-narratological-approach-to-content-selection-in-a-Vercauteren/fea57bb43caf25c12199ee38335bb7b7e0e04438.