SilencioNetwork/african-languages-speech

🌍 African Languages Speech Dataset

---

🌍 Underrepresented languages, finally represented.

This dataset provides high-quality voice data for African languages that are severely underrepresented in existing speech corpora.

📧 Need more? sofia@silencioai.com — we have 25,000+ hours of African language data available.

---

🎯 Dataset Overview

347 recordings across 6 African languages, sourced from native speakers across the continent.

Language	Code	Samples	Region	Speakers
🇳🇬 Nigerian English	english_nigeria	75	Nigeria	Native Nigerian speakers
🇰🇪 Swahili	swahili	70	East Africa	Kenya, Tanzania
🇪🇹 Amharic	amharic	53	Ethiopia	Native speakers
🇳🇬 Yoruba	yoruba	50	Nigeria	Southwest Nigeria
🇳🇬 Igbo	igbo	50	Nigeria	Southeast Nigeria
🇳🇬 Hausa	hausa	49	West Africa	Nigeria, Niger, Ghana
Total		347

🚀 Quick Start

from datasets import load_dataset

# Load Swahili samples
swahili = load_dataset("SilencioNetwork/african-languages-speech", "swahili")

# Load Nigerian languages
hausa = load_dataset("SilencioNetwork/african-languages-speech", "hausa")
yoruba = load_dataset("SilencioNetwork/african-languages-speech", "yoruba")
igbo = load_dataset("SilencioNetwork/african-languages-speech", "igbo")

# Load Nigerian English (African-accented English)
nigerian_english = load_dataset("SilencioNetwork/african-languages-speech", "english_nigeria")

# Process samples
for sample in swahili['train']:
    audio = sample['audio']
    transcript = sample['transcript']
    print(f"Transcript: {transcript[:50]}...")

🌍 Why African Languages?

Most speech AI is trained on English and European languages. African languages are massively underrepresented:

• 🗣️ 2,000+ languages spoken across Africa
• 📊 <1% representation in major speech datasets
• 🚀 Huge demand for ASR/TTS in local languages
• 💰 Growing market — African tech is booming

This dataset helps bridge that gap.

📊 Full Data Availability

This sample is <2% of our African language corpus.

Language	Sample	Full Corpus Available
Swahili	70	9,700+ hours
Nigerian English	75	8,100+ hours
Hausa	49	2,500+ hours
Yoruba	50	1,800+ hours
Igbo	50	1,200+ hours
Amharic	53	900+ hours
Other African	—	5,000+ hours
Total	347	25,000+ hours

📋 Metadata

Each recording includes:

Field	Description
file_name	Path to audio file
id	Unique recording ID
gender	Speaker gender
country_code	ISO country code
mother_tongue	Native language
dialect	Regional dialect
duration	Recording length (seconds)
emotions	Emotion labels
type_of_script	free_speech / keywords / monologues
transcript	Whisper-generated transcription
script	Original prompt

🎤 Audio Format

• Format: WAV
• Sample Rate: 48 kHz
• Recording: Real-world conditions (mobile devices, natural environments)

🎯 Use Cases

• 🗣️ African language ASR — Build speech recognition for underserved markets
• 🔊 TTS for local languages — Voice assistants, accessibility tools
• 🌍 Multilingual models — Improve coverage for African languages
• 📱 Voice apps for Africa — Banking, healthcare, education
• 📊 Benchmarking — Test model performance on low-resource languages

⚖️ License

CC BY-NC 4.0 — Free for research and non-commercial use.

For commercial licensing, contact sofia@silencioai.com.

📧 Get More Data

Building voice AI for Africa? We can help.

What You Need	We Provide
More Swahili	✅ 9,700+ hours
More Hausa/Yoruba/Igbo	✅ 5,500+ hours combined
Other African languages	✅ Zulu, Xhosa, Twi, Wolof, more
Custom collection	✅ We can source any language

📧 Email: sofia@silencioai.com

🌐 Website: silencioai.com

---

Citation

@dataset{silencio_african_speech_2025,
    title        = {African Languages Speech Dataset},
    author       = {Silencio Network},
    year         = {2025},
    publisher    = {Hugging Face},
    license      = {CC BY-NC 4.0}
}