Audio-Based Understanding of Audiobook Narration Appeal

This study investigates how vocal and acoustic features influence audiobook appeal by analyzing LibriVox data. It establishes a robust association between narration qualities and consumption metrics, even after accounting for title effects.

• Extracted tone, pace, and loudness from LibriVox using pre-trained audio models.
• Analyzed the relationship between these features and view-rate consumption data.
• Validated findings using proprietary engagement metrics to ensure nuance.
• Identified interplay between narration qualities, genre, and title.

The authors consider this significant as the first systematic computational study linking these factors, highlighting potential for improved audiobook personalization and narrator casting.