Sprecher
Beschreibung
ABSTRACT
Since 2024, the advent of high-quality AI-generated voices – driven by advances in speech synthesis technologies – has sparked widespread public discourse in German newspapers and feuilletons. The potential integration of artificial voices into the audiobook market raises fundamental questions concerning listener perception, notions of authenticity, and the broader cultural acceptance of machine-generated narrators as cost-efficient alternatives to professional voice actors and narrators. This study investigates whether AI-generated and human audiobook narrators are evaluated differently by listeners, and whether these evaluations are systematically influenced by contextual and individual variables. Specifically, it examines the effects of the actual source of the voice (human vs. AI), the belief about the voice’s origin (informed AI vs. uninformed), textual features such as direct speech dialogues or scenic narration, and individual differences in technological affinity. To investigate these questions, an online experiment (N = 500) was conducted using a 2x2 between-subjects design. The study manipulated both the type of voice (AI vs. human) and the information participants received about the voice’s origin (told it is AI vs. not told). Self-identified audiobook listeners were randomly assigned to one of four experimental conditions. Each participant listened to four one-minute audiobook excerpts, evaluating the narrator’s voice after each stimulus presentation. The stimuli included professionally produced audiobook passages and corresponding AI-generated versions created with the current state-of-the-art voice synthesis software from ElevenLabs. As one of the first empirical investigations of AI voice evaluation in the German audiobook domain, this study provides foundational insights into how AI-voices are evaluated by listeners. The findings will provide valuable insights into how generated voices are perceived and under which conditions they may (or may not) be accepted as narrators of literary texts. The findings offer a critical contribution to emerging research on human-AI interaction in cultural contexts and highlight the need for nuanced understanding as AI-generated voices increasingly permeate everyday media consumption.
BIOGRAPHICAL NOTES
Dr. Annika Schwabe, MSc BSc BA is a research associate (post-doc) at the University of Vienna as part of the Listening to Literature: Experiencing Literary Audiobooks project. She wrote her dissertation in Media Psychology on cognitive and emotional aspects of the digitization of fiction reading, which she defended in November 2023. She also studied Psychology (BSc & MSc) at the University of Vienna with a focus on Research Methods and Statistics and holds an additional degree in Development Studies (BA) from the University of Vienna.
Mag. Dr. Lukas Kosch is a research assistant and lecturer at the Department of German Studies at the University of Vienna, currently working on the project Listening to Literature: Experiencing Literary Audiobooks. After studying German language and literature and history, he focused on post-war philosophy, the consequences of the digitalization of literature and the processes of literary reading from a reception theory perspective. His most recent publications are the anthology “Mythen des Lesens: Über eine Kulturtechnik in Zeiten gesellschaftlichen Wandels” (transcript 2024) and in Mai 2025 the monograph “Literarisches Lesen: Von der literaturwissenschaftlichen Lesetheorie zur transdisziplinären Leseforschung” (Wallstein) will be published.
https://listening-to-literature.univie.ac.at/en/