SilencioNetwork/yoruba-speech

Yoruba Speech Dataset

The most comprehensive Yoruba speech dataset on HuggingFace - natural, real-world Yoruba from native speakers in Nigeria and the diaspora.

Dataset Overview

• Total audio samples: 39 recordings
• Total duration: ~22 minutes
• Primary region: Nigeria (Southwest - Ibadan, Lagos)
• Context: Natural spontaneous speech (free_speech)
• Audio format: WAV files
• Sample rate: 48 kHz
• License: CC BY-NC 4.0 (free for research, non-commercial use)

Language Context

Yoruba (Èdè Yorùbá) is one of Africa's major languages:

• Speakers: 45M+ native speakers
• Geographic spread: Nigeria (Southwest), Benin, Togo, diaspora (UK, US, Brazil)
• Tonal language: 3 tones (high, mid, low) - essential for meaning
• Niger-Congo family: Closely related to Igbo, Edo
• Cultural significance: Rich oral literature, proverbs, music tradition (Afrobeat)
• Digital presence: Growing use in social media, YouTube, voice apps

Target Applications

This dataset is designed for:

• Yoruba ASR systems - Speech recognition for 45M+ speakers
• Voice assistants - Banking, customer service, government services in Nigeria
• TTS for Yoruba - Text-to-speech with authentic Nigerian pronunciation
• Language learning apps - Pronunciation training (including tonal patterns)
• Content moderation - Social media platforms operating in Nigeria
• Cultural preservation - Digitizing Yoruba oral traditions, music, stories

Dataset Structure

yoruba-speech/
└── data/
    ├── audio/              # 39 WAV files
    └── metadata.csv        # Speaker metadata & transcripts

Data Splits

Yoruba (Nigeria)

• Files: 39 recordings
• Dialect: Primarily Standard Yoruba (Ibadan variant)
• Context: Natural spontaneous speech
• Use case: General-purpose Yoruba ASR, Nigerian voice AI

Languages Sampled in This Dataset ✅

39 audio samples available for immediate download:

• Yoruba: 39 files (~22 minutes)

Full OTS Inventory Available 📊

This sample represents <0.08% of Silencio's complete Yoruba speech inventory.

Contact us for access to our full Yoruba corpus:

Yoruba by Country:

• Nigeria: 1,884 hours, 220,362 recordings
• United States: 9 hours, 977 recordings
• Benin: 6 hours, 875 recordings
• United Kingdom: 3 hours, 335 recordings
• Ghana: 3 hours, 403 recordings
• Kenya: 2 hours, 295 recordings
• American Samoa: 1 hour, 191 recordings
• Niger: 1 hour, 173 recordings
• Andorra: 1 hour, 161 recordings
• + 15 more countries (diaspora communities)

Total: 1,917+ hours across 224,000+ recordings

Contact us for access: sofia@silencioai.com

Key Features

✅ Native speakers - Authentic Nigerian Yoruba (Southwest region)
✅ Natural speech - Real conversational Yoruba, not scripted
✅ Tonal language - Captures high, mid, low tone distinctions
✅ Diverse topics - Daily life, opinions, cultural topics
✅ Standard dialect - Ibadan variant (widely understood)
✅ High audio quality - 48 kHz WAV format
✅ Rich metadata - Gender, dialect, emotions, transcriptions
✅ Ethical data collection - Consent-based, privacy-preserving

Use Cases

1. Yoruba Speech Recognition

Build ASR systems for the 45M+ Yoruba-speaking market in Nigeria, Benin, Togo, and the diaspora.

2. Voice Banking & Fintech

Power voice-enabled banking apps and financial services in Southwest Nigeria (Lagos, Ibadan, Abeokuta).

3. Yoruba TTS

Train text-to-speech models with authentic Nigerian Yoruba pronunciation and tonal patterns.

4. Content Moderation

Build speech detection for Nigerian social media platforms (Nairaland, local Facebook groups).

5. Language Learning

Develop pronunciation training tools for Yoruba learners (especially tone recognition).

6. Cultural Preservation

Digitize Yoruba oral traditions, proverbs (òwe), folk stories (ìtàn), and music.

Loading the Dataset

from datasets import load_dataset

# Load full Yoruba dataset
dataset = load_dataset("SilencioNetwork/yoruba-speech")

# Access samples
for sample in dataset['train']:
    audio = sample['audio']
    transcript = sample['transcript']
    dialect = sample['dialect']
    print(f"Transcript: {transcript}")
    print(f"Dialect: {dialect}")

Sample Metadata

Each recording includes:

• file_name: Audio file path
• id: Unique recording ID
• gender: Speaker gender
• location: Speaker location
• mother_tongue: Native language (Yoruba)
• dialect: Regional variant (Nigeria - Ibadan)
• duration: Recording length (seconds)
• emotions: Emotion labels (focused, relaxed, excited, etc.)
• language: Yoruba
• type_of_script: free_speech (spontaneous, unscripted)
• transcript: Whisper-generated transcription (Yoruba text)
• script: Original prompt (question asked in Yoruba)

Yoruba Speech Characteristics

This dataset captures authentic Yoruba speech features:

• Tonal phonology: 3-tone system (high, mid, low) - critical for word meaning
• Vowel harmony: ATR (advanced tongue root) harmony patterns
• Nasal consonants: Distinctive nasalization
• Syllable structure: Primarily CV (consonant-vowel)
• Natural prosody: Authentic rhythm, stress, intonation
• Real-world audio: Mobile recordings, natural environments

Market Context

Nigerian Tech & Economy

• 45M+ Yoruba speakers - One of Nigeria's 3 major languages
• Southwest Nigeria: Economic powerhouse (Lagos = 24M population)
• Lagos GDP: $136B (larger than 30+ African countries)
• Growing smartphone penetration: 60%+ in Southwest Nigeria
• Digital payment revolution: Voice commands for fintech emerging
• YouTube/TikTok: Growing Yoruba content ecosystem

Why Yoruba Matters

• Underrepresented in AI: <0.1% of speech datasets despite 45M+ speakers
• High commercial value: Banking, telecom, e-commerce in Southwest Nigeria
• Cultural significance: Rich oral tradition, Afrobeat music (Fela Kuti, Burna Boy)
• Government use: Lagos State increasingly using Yoruba in public services
• Diaspora market: Large communities in UK, US, Brazil

Tonal Language Considerations

Yoruba is a tonal language - pitch changes word meaning:

• oko (high-mid) = husband
• òkò (mid-low) = hoe (farming tool)
• ọkọ̀ (mid-low with mid vowel) = vehicle

ASR/TTS systems need to capture these tone distinctions for accurate Yoruba processing.

Ethical Considerations

All data was collected with explicit informed consent from native Yoruba speakers. Recordings contain general conversational topics only - no sensitive personal information.

Comparison to Other Datasets

Dataset	Language	Hours	Speakers	Natural?
LibriSpeech	English	1,000	2,484	❌ Read speech
Common Voice	Yoruba	~5	Few	⚠️ Read sentences
Silencio Yoruba	Yoruba	1,917+	10,000+	✅ Spontaneous

This is the largest natural Yoruba speech dataset available.

Citation

If you use this dataset in your research or commercial product, please cite:

@dataset{silencio_yoruba_speech_2026,
  title={Yoruba Speech Dataset},
  author={Silencio Network},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/SilencioNetwork/yoruba-speech}
}

Related Datasets

• African Languages Speech - 6 African languages (Swahili, Hausa, Yoruba, Igbo, Amharic, Nigerian English)
• Complete Voice AI Speech Dataset - 39 language/accent variants
• Indian Languages Speech - 9 Indian languages
• Global English Accents Speech - 20 English accent variants

License

CC BY-NC 4.0 (Creative Commons Attribution-NonCommercial 4.0 International)

✅ Free for research and non-commercial use
❌ Commercial use requires licensing (contact us)

About Silencio

Silencio is a voice AI data sourcing company with 2M+ contributors across 180+ countries. We provide scaled sourcing of real-world audio and speech data for AI labs, robotics companies, and enterprises building voice AI products.

🌐 silencioai.com
📧 sofia@silencioai.com

---

Tags: yoruba, edé yorùbá, nigerian languages, west africa, nigeria, tonal language, african languages, low-resource languages, speech recognition, asr, tts, voice ai, natural speech, spontaneous speech, nigerian speech, lagos, ibadan