LJ Speech single-speaker English TTS dataset
- Provider
- Keith Ito, Linda Johnson
- License
- Public Domain (source)
- Size
- 10 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/MikhailT/lj-speech)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Speech & Audio
- Source
- https://keithito.com/LJ-Speech-Dataset/
About LJ Speech single-speaker English TTS dataset
The LJ Speech Dataset is a public domain speech dataset of 13,100 short audio clips of a single speaker reading passages from 7 non-fiction books, with a transcription for each clip. Clips vary from 1 to 10 seconds and total approximately 24 hours; the audio was recorded in 2016-17 by Linda Johnson for LibriVox and the texts were published between 1884 and 1964. Released by Keith Ito in 2017, it became the default benchmark for single-speaker neural text-to-speech and vocoders (Tacotron 2, FastSpeech, WaveGlow, HiFi-GAN, VITS all report on it). This repository holds the full dataset as one split named 'full' in 8 Parquet shards under data/ (measured 2026-11-14); the card lists 13,100 examples at 22,050 Hz, with columns for the file name, the spoken text, the normalised text and the audio. Load with datasets.load_dataset('MikhailT/lj-speech') or read the shards with pyarrow or pandas. The original publisher states the dataset is in the public domain in the USA; the upload's 'cc-by-4.0' tag is the uploader's and the upstream public domain status governs. Credit Keith Ito and Linda Johnson when using it.
How to load LJ Speech single-speaker English TTS dataset in Python
from datasets import load_dataset
ds = load_dataset("MikhailT/lj-speech")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: data/full-00000-of-00008-0cc4e31b6b21f574.parquet
Analyze this dataset without downloading it
On Scigantic, LJ Speech single-speaker English TTS dataset opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Speech & Audio datasets
- YODAS2: YouTube-Oriented Dataset for Audio and Speech (long-form)
- LibriTTS-R: restored LibriTTS speech corpus
- LibriTTS multi-speaker English text-to-speech corpus
- LibriSpeech-Long: Long-Form Speech Benchmark of Contiguous Audiobook Segments up to 4 Minutes (Google DeepMind)
- CML-TTS multilingual speech synthesis dataset
- Omnilingual Asr Corpus
- VoiceBank+DEMAND noisy speech database (16 kHz)
- Bach Doodle Dataset: 21.6 Million User Melodies with Coconet Four-Voice Bach-Style Harmonizations (Google Magenta)