scripts/prepare_data.py

#!/usr/bin/env python3
"""
Prepare Nigerian language audio data for XTTS fine-tuning.
Supports: Yoruba (yo), Hausa (ha), Igbo (ig), Pidgin (pcm)
"""
import os
import json
import zipfile
from pathlib import Path

BASE_DIR = Path.home() / "voice-training"
DATASETS_DIR = BASE_DIR / "datasets"
OUTPUT_DIR = BASE_DIR / "prepared_data"

def extract_openslr86():
    """Extract OpenSLR86 Yoruba dataset."""
    print("=== Extracting OpenSLR86 Yoruba ===")
    female_zip = DATASETS_DIR / "yo_ng_female.zip"
    male_zip = DATASETS_DIR / "yo_ng_male.zip"
    
    yoruba_dir = OUTPUT_DIR / "yoruba"
    yoruba_dir.mkdir(parents=True, exist_ok=True)
    
    for zip_path in [female_zip, male_zip]:
        if zip_path.exists():
            print(f"Extracting {zip_path.name}...")
            with zipfile.ZipFile(zip_path, 'r') as zf:
                zf.extractall(yoruba_dir)
            print(f"Extracted {zip_path.name}")
        else:
            print(f"Waiting for {zip_path.name} to download...")
    
    return yoruba_dir

def prepare_manifest(audio_dir: Path, output_file: Path, lang: str):
    """Create a training manifest file."""
    manifest = []
    
    # Look for audio files
    for ext in ['*.wav', '*.mp3', '*.flac']:
        for audio_file in audio_dir.rglob(ext):
            # Look for corresponding text file
            text_file = audio_file.with_suffix('.txt')
            if text_file.exists():
                text = text_file.read_text().strip()
            else:
                text = ""
            
            manifest.append({
                "audio_file": str(audio_file),
                "text": text,
                "language": lang,
                "duration": 0  # Will calculate later
            })
    
    print(f"Found {len(manifest)} audio files for {lang}")
    
    # Save manifest
    output_file.parent.mkdir(parents=True, exist_ok=True)
    with open(output_file, 'w') as f:
        json.dump(manifest, f, indent=2, ensure_ascii=False)
    
    return len(manifest)

def main():
    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
    
    print("Preparing Nigerian language datasets for TTS training...")
    print(f"Output directory: {OUTPUT_DIR}")
    
    # Extract OpenSLR86 if downloaded
    if (DATASETS_DIR / "yo_ng_female.zip").exists():
        yoruba_dir = extract_openslr86()
        prepare_manifest(yoruba_dir, OUTPUT_DIR / "manifests" / "yoruba.json", "yo")
    
    print("\n=== Dataset preparation complete ===")
    print(f"Check {OUTPUT_DIR} for prepared data")

if __name__ == "__main__":
    main()