AI aims to make character voices more distinct in audiobooks
The Audible team described a system that assigns consistent voices to characters and directs their delivery based on the story. Tests on English and Spanish books produced a better listening experience.
Listeners need to know who is speaking. Amazon Science describes how the Audible team combines text-to-speech with analysis of the entire story. According to the authors, current general-purpose voice models struggle to maintain distinct character voices in long books. The test included mysteries, science fiction and romance novels. Dialogue differentiation improved by 3.7 points in the MUSHRA listening test, in which people rate audio. Main characters receive distinct profiles from a prepared voice catalogue, while minor characters may retain the narrator’s voice.
The BookLLM system correctly assigned the speaker in 97 % of English cases and 94 % of Spanish cases in the test. The authors tested 25 books. These are findings from the developer’s research, rather than an announcement of a feature available to all Audible listeners. Two human evaluators prepared the reference answers for comparison. Short instructions for individual lines specify delivery such as nervous whispering or angry shouting. Amazon Science.