scripts/tts-kokoro.py

import argparse, os, sys, wave, struct, json, math, re
from pathlib import Path

parser = argparse.ArgumentParser()
parser.add_argument('--text-file', required=True)
parser.add_argument('--output', required=True)
args = parser.parse_args()

with open(args.text_file, 'r', encoding='utf-8') as f:
    text = f.read().strip()

if not text:
    print("ERROR: Empty text file")
    sys.exit(1)

print(f"  Text: {len(text)} chars")
print(f"  Output: {args.output}")

# Placeholder: Kokoro-82M integration
# kokoro = Kokoro("hexgrad/Kokoro-82M")
# audio = kokoro.generate(text)
# audio.export(args.output, format="wav")

# For now, check if Kokoro package is available
try:
    from kokoro import KPipeline
    pipeline = KPipeline(lang_code='a')  # 'a' for American English / auto
    generator = pipeline(text, voice='af_heart')
    audio_chunks = []
    for i, (gs, ps, audio) in enumerate(generator):
        audio_chunks.append(audio)
    if audio_chunks:
        import numpy as np
        import soundfile as sf
        full_audio = np.concatenate(audio_chunks)
        sf.write(args.output, full_audio, 24000)
        print(f"  -> Generated TTS: {args.output}")
    else:
        print("  WARNING: No audio generated")
except ImportError:
    print("  WARNING: kokoro package not installed. Install with:")
    print("    pip install kokoro>=0.9.2 soundfile")
    print("  Creating placeholder...")
    os.makedirs(os.path.dirname(args.output), exist_ok=True)
    sample_rate = 24000
    duration = int(len(text) * 0.15)
    with wave.open(args.output, 'w') as wf:
        wf.setnchannels(1)
        wf.setsampwidth(2)
        wf.setframerate(sample_rate)
        for i in range(sample_rate * duration):
            t = i / sample_rate
            val = int(16000 * math.sin(2 * math.pi * 220 * t) * max(0, 1 - t/duration))
            wf.writeframes(struct.pack('<h', val))
    print(f"  -> Placeholder TTS: {args.output} ({duration}s)")