Skip to content

Dataset · Alex-Song

Multilingual speech recognition corpus (86,300 hours)

Alex-Song/MSR-86K

A large 15-language YouTube audio corpus for training and researching speech recognition systems.

What's inside

A large-scale evolving ASR dataset containing 86,300 hours of transcribed audio. The recordings are sourced from public YouTube videos and cover 15 languages, including English, Spanish, German, French, Korean, and Hindi.

What you can do with it

  • Training and fine-tuning multilingual speech recognition (ASR) models
  • Researching transcription algorithms for spontaneous speech from video
Tasks
Automatic speech recognition
Languages
Spanish, Korean, English, French, German, Hindi, Vietnamese, Italian and 4 more
Downloads this month
344
Likes
60
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("Alex-Song/MSR-86K")