# Hausa TTS NaijaVoices 300 - WAV Files Dataset

This dataset contains **300 decodable WAV audio files** for Hausa TTS training.

## Files

- `audio/`: Folder containing `.wav` audio files (24kHz)
- `metadata.csv`: CSV with file_name, text, language, duration
- `metadata.jsonl`: JSONL with the same metadata

## Audio Properties

- Format: WAV (PCM)
- Sample Rate: 24kHz
- Duration: 2-15 seconds (verified)
- Channels: Mono

## Usage

```python
import pandas as pd
import soundfile as sf

metadata = pd.read_csv('metadata.csv')
for _, row in metadata.iterrows():
    audio, sr = sf.read(row['file_name'])
    text = row['text']
    # Use audio and text for TTS training
```

## Dataset Statistics

- Samples: 300
- Min duration: 2.04s
- Max duration: 7.86s
- Mean duration: 3.09s

## Source

Derived from `vaghawan/hausa-tts-24khz-naijavoices-O0456`.