Inteligencia artificial en la audiodescripción: traducción accesible de contenidos culturales

Introduction: The aim of this research is to reflect on the challenges and limitations arising from the use of artificial intelligence in the accessible translation of cultural objects, particularly sculptures, and in the production of audio descriptions for an audience with visual impairments. Methods: Based on a semiotic analysis of the artworks, this research proposes, on the one hand, a version created by the authors of this study following the guidelines of audio description, and on the other hand, a second version produced with the assistance of artificial intelligence. The obtained audio descriptions have been analyzed using a platform that has been trained for this purpose. Results: The analysis of the audio descriptions generated by the artificial intelligence systems UniDescription and Microsoft Copilot enabled the identification of relevant proposals in terms of textual organization and linguistic choices. Conclusions: The study reveals that the combination of human experience and artificial intelligence–assisted translation tools can effectively improve the quality and efficiency of accessible content creation services.