Dataset · facebook
Multilingual LibriSpeech (MLS) for Speech
facebook/multilingual_librispeech
LibriVox audiobooks in 8 languages with texts for ASR, TTS, and speech research.
What's inside
This is a streamable version of Multilingual LibriSpeech: archives are rebuilt from original OpenSLR data for convenient streaming. The corpus consists of read LibriVox audiobooks in English, German, Dutch, Spanish, French, Italian, Portuguese, and Polish. Modalities include audio and text, with a metadata volume between 1M and 10M.
What you can do with it
- Training and evaluating multilingual ASR models
- Training TTS on audiobooks
- Speech research with text transcripts
- Tasks
- Automatic speech recognition, Text → speech, Text → audio
- Languages
- German, Dutch, French, Italian, Spanish, Portuguese, Polish, English
- Size
- 1M–10M rows
- Downloads this month
- 38.9K
- Likes
- 201
Models trained on it
- neuphonic/neucodecAudio → audio 55K
- parler-tts/parler-tts-mini-multilingual-v1.1Text → speech 12.6K
- nvidia/stt_en_conformer_ctc_largeAutomatic speech recognition 7.3K
- bofenghuang/asr-wav2vec2-ctc-frenchAutomatic speech recognition 5.9K
- OuteAI/OuteTTS-0.2-500M-GGUFText → speech 2.9K
- pnnbao-ump/VieNeu-TTS-v2-Turbo-GGUFText → speech 2.8K
- alien79/f5-ita-testText → speech 2.2K
- bofenghuang/whisper-large-v3-frenchAutomatic speech recognition 1.8K
How to load it
from datasets import load_dataset
ds = load_dataset("facebook/multilingual_librispeech")