Dataset · mythicinfinity
Multispeaker English speech corpus for synthesis
mythicinfinity/libritts
About 585 hours of studio English speech and texts for training text-to-speech models.
What's inside
Contains pairs of 24 kHz audio recordings and English text transcripts. The dataset size is about 585 hours (between 100k and 1M records) from multiple speakers. The corpus was created with Google Speech and Google Brain teams based on LibriVox audiobooks and Project Gutenberg texts from the LibriSpeech project.
What you can do with it
- Training multi-speaker text-to-speech (TTS) models
- Fine-tuning and voice cloning in English
- Testing 24 kHz speech generation quality
- Tasks
- Text → speech
- Languages
- English
- Size
- 100K–1M rows
- Downloads this month
- 7.7K
- Likes
- 34
Models trained on it
- lucadellalib/focalcodec_50hzAudio → audio 1.2K
- lucadellalib/focalcodec_12_5hzAudio → audio 888
- lucadellalib/focalcodec_25hzAudio → audio 503
- ibm-research/DAC.speech.v1.0 129
- hr16/multilingual_spinEmbeddings 124
- prakharkhatri/IndiCodecAudio → audio 101
- lucadellalib/dycastAudio → audio 69
- rohansheth/tiro-qwen3-asr-1.7b-ffasr-v1Automatic speech recognition 36
How to load it
from datasets import load_dataset
ds = load_dataset("mythicinfinity/libritts")