Skip to content

Dataset · speechbrain

Speech audio recordings in 45 languages for language identification

speechbrain/common_language

45.1 hours of speech audio recordings from CommonVoice (one hour per language) for training voice language recognition systems.

What's inside

The dataset contains speech audio recordings selected from the open CommonVoice base by the SpeechBrain team. Total duration is 45.1 hours—exactly one hour of material for each of the represented languages, including Arabic, Breton, Catalan, Czech, and Chuvash.

What you can do with it

  • Training classification models and automatic language identification by audio (Language ID)
  • Testing multilingual speech encoders on rare and low-resource languages
Tasks
Audio classification
Languages
Arabic, Breton, Catalan, cnh, Czech, Chuvash, Welsh, German and 4 more
Size
100K–1M rows
Downloads this month
532
Likes
45

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("speechbrain/common_language")