license:
- cc-by-4.0
task_categories:
- text-to-speech
language:
- en
pretty_name: ame
size_categories:
- 10M<n<100M
configs:
- config_name: default
data_files:
- split: dev
path: data/dev-*
- split: test
path: data/test-*
- split: train
path: data/train-*
Aligned English Multilingual LibriSpeech (MLS) Dataset
This repository contains phoneme alignments for 44.5K hours of English audio from the Multilingual LibriSpeech (MLS) dataset, generated using the Montreal Forced Aligner. The aligned data provides precise timing information for phonemes in each audio file, making it particularly valuable for training Text-to-Speech (TTS) models that require accurate phoneme duration estimation.
Dataset Contents
Our repository provides the TextGrid alignment files only. For the corresponding audio files, please download the Parler MLS English dataset from their Hugging Face repository: parler-tts/mls_eng.
Getting Started
Downloading the Dataset
To download the aligned TextGrid files, use the Hugging Face CLI:
huggingface-cli download ntt123/aligned_mls_eng --repo-type dataset --local-dir aligned_mls_eng
Working with TextGrid Files
The dataset uses TextGrid files to store alignment information. To read these files, you'll need the textgrid Python library:
- Install the library:
pip install textgrid
- Load and process TextGrid files:
import textgrid
# Load a TextGrid file
tg = textgrid.TextGrid.fromFile('test.TextGrid')