LibriTTS-R

A sound-quality-restored version of LibriTTS — 585 hours of 24 kHz English read speech from 2456 speakers, identical samples/texts to LibriTTS but enhanced via Google's Miipher speech restoration model.

• Task: tts
• Languages: en
• Hours: 585
• Domain: audiobook (restored, multi-speaker read)
• License: CC BY 4.0
• Homepage: https://www.openslr.org/141/
• Paper: https://arxiv.org/abs/2305.18802

Recommendation

The current default large-scale multi-speaker English TTS corpus — pick over plain LibriTTS whenever audio fidelity matters (zero-shot TTS, neural codec training, voice cloning baselines). Drop-in compatible with any LibriTTS pipeline; same speaker IDs, splits, and transcripts.

Getting the data

Downloadable via VoxKitchen (libritts_r, source: openslr, size: 930 MB - 43.6 GB):

vkit docker download --tag slim libritts_r --root ./data/libritts_r

Subsets: dev-clean, dev-other, test-clean, test-other, train-clean-100, train-clean-360, train-other-500.

Still derived from LibriVox audiobooks, so prosody is narrative/literary; expect domain shift if your target is conversational.

Suggested processing

A recommended VoxKitchen pipeline ships in the repository at voxkitchen/templates/pipelines/tts-data-prep.yaml — run it with vkit docker run.