The dataset viewer is not available for this split.
Parquet error: Scan size limit exceeded: attempted to read 606892899 bytes, limit is 300000000 bytes
Make sure that
1. the Parquet files contain a page index to enable random access without loading entire row groups2. otherwise use smaller row-group sizes when serializing the Parquet files
Error code: TooBigContentError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
🇮🇳 South Asian Languages Speech Dataset
🌏 Voice data for 2 billion+ South Asian speakers.
📧 Need more? sofia@silencioai.com — we have 15,000+ hours of South Asian language data.
🎯 Dataset Overview
137 recordings across 3 major South Asian languages.
| Language | Code | Samples | Region |
|---|---|---|---|
| 🇵🇰 Urdu | urdu |
51 | Pakistan, India |
| 🇧🇩 Bengali | bengali |
45 | Bangladesh, India |
| 🇮🇳 Hindi | hindi |
41 | India |
🚀 Quick Start
from datasets import load_dataset
# Load Hindi
hindi = load_dataset("SilencioNetwork/south-asian-speech", "hindi")
# Load Urdu
urdu = load_dataset("SilencioNetwork/south-asian-speech", "urdu")
# Load Bengali
bengali = load_dataset("SilencioNetwork/south-asian-speech", "bengali")
for sample in hindi['train']:
audio = sample['audio']
transcript = sample['transcript']
print(f"Hindi: {transcript[:50]}...")
🌏 Why South Asian Languages?
South Asia is home to 2 billion+ people speaking hundreds of languages:
- 🇮🇳 Hindi — 600M+ speakers, India's most spoken language
- 🇵🇰 Urdu — 230M+ speakers, Pakistan's national language
- 🇧🇩 Bengali — 270M+ speakers, Bangladesh + West Bengal
Yet these languages remain underrepresented in most speech datasets.
📊 Full Data Availability
| Language | Sample | Full Corpus |
|---|---|---|
| Urdu | 51 | 5,900+ hours |
| Hindi | 41 | 3,800+ hours |
| Bengali | 45 | 2,100+ hours |
| Tamil | — | 1,500+ hours |
| Telugu | — | 1,200+ hours |
| Marathi | — | 800+ hours |
| Total South Asian | 137 | 15,000+ hours |
🎯 Use Cases
- 🗣️ Indic ASR — Speech recognition for Hindi, Urdu, Bengali
- 🔊 TTS — Voice synthesis for South Asian markets
- 📱 Voice assistants — Siri, Alexa, Google for India/Pakistan
- 🏥 Healthcare — Voice AI for regional healthcare
- 📞 Call centers — Automated support in local languages
⚖️ License
CC BY-NC 4.0 — Free for research and non-commercial use.
For commercial licensing, contact sofia@silencioai.com.
📧 Get More Data
📧 Email: sofia@silencioai.com
🌐 Website: silencioai.com
Citation
@dataset{silencio_south_asian_2025,
title = {South Asian Languages Speech Dataset},
author = {Silencio Network},
year = {2025},
publisher = {Hugging Face},
license = {CC BY-NC 4.0}
}
- Downloads last month
- 5