Skip to content

Dataset · voice-is-cool

VoxTube multilingual speaker dataset

voice-is-cool/voxtube

A multilingual speech recognition dataset from YouTube videos with over 4 million utterances.

What's inside

VoxTube contains audio signals and metadata (speaker IDs, sessions, languages) collected from YouTube videos under the CC BY 4.0 license. It includes 5,040 unique speakers and about 4 million utterances across more than 10 languages, including en, ru, es, pt, fr, ar.

What you can do with it

  • Training and evaluation of automatic speech recognition (ASR) systems
  • Training speaker identification and verification (Speaker Recognition) models
  • Research in multilingual audio analysis
Tasks
Audio classification
Languages
English, Russian, Spanish, Portuguese, French, Arabic, Italian, German and 3 more
Size
1M–10M rows
Downloads this month
748
Likes
20
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("voice-is-cool/voxtube")