There are four audiobooks books in this dataset.

Identifier	Book
emma	Emma, by Jane Austen
huck	The Adventures of Huckleberry Finn, by Mark Twain
sherlock	The Adventures of Sherlock Holmes, by Arthur Conan Doyle
treasure	Treasure Island, by Robert Louis Stevenson

The data is provided in two formats: chapter-level and utterance-level.

Each book is read by multiple speakers. Waveforms are single-channel 16-bit PCM WAVs sampled at 22050 Hz. Note that the waveforms were original provided by LibriVox as MP3 files.

Chapter Data

This format of the data contains the audiobook recording at the chapter-level. Each chapter-level waveform is accompanied by the text and its respective word-level alignment. Alignment was performed using the methods described in [1].

Data type	Description
./{book}/wav/{speaker}/.wav	Waveforms for book chapters
./{book}/txt/.txt	Text files with the corresponding text for book chapters
./{book}/ctm/{speaker}*.ctm	ctm text file with word-level alignment.

Segmented Data

The chapter-level alignment was used to segment the data into utterances. This provides a more convenient way to use the data for Text-to-Speech synthesis or Voice Conversion. Waveforms are organized by speaker and utterance identifiers are consistent across speakers within each book. If utterances are missing for a given speaker, it is implied that the alignment failed, so the utterance was excluded.

Some books are read jointly by multiple speakers (denoted various in the chapter-level data). For this reason, some speakers may have a very small amount of data. Note that speaker names are consistent across the entire dataset, so additional data for a single speaker may be acquired by pooling recordings from different books.

Data type	Description
./{book}/wav/{speaker}/*.wav	Waveforms following chpXX_uttXXX naming convention.
txt.punc	Original text without any processing other than sentence tokenization.
txt.clean	Cleaned text stripped of punctuation and (very minor) text normalization
spk2gender	Table mapping speakers to gender
spk2utt	Table mapping speakers to a list of available utterances
utt2spk	Table mapping utterances to a list of speakers
spk2dur	Table mapping speakers to amount of data in hours