src/pro_audio_generator.py

"""
Pro Audio Generator — Edge TTS (instant, no GPU/RAM needed).
Replaces XTTS-v2 to prevent OOM crashes on HF Spaces.
"""
import os
import asyncio
import time
import edge_tts

# English male voice — deep and engaging
VOICE = os.environ.get("TTS_VOICE", "en-US-GuyNeural")


def process_script_to_audio(script_text, output_filename="output.mp3"):
    """
    Generates audio from text using Edge TTS.
    Instant generation, no model loading, no RAM usage.
    Returns path to generated audio file.
    """
    output_dir = "output/audio"
    os.makedirs(output_dir, exist_ok=True)

    # Ensure .mp3 extension
    if output_filename.endswith(".wav"):
        output_filename = output_filename.replace(".wav", ".mp3")

    output_path = os.path.join(output_dir, output_filename)

    print(f"🎙️ Generating Audio ({len(script_text)} chars) with Edge TTS...", flush=True)
    start_time = time.time()

    try:
        # Edge TTS is async, run it
        asyncio.run(
            edge_tts.Communicate(
                script_text,
                VOICE,
                rate="+5%"  # Slightly faster for engagement
            ).save(output_path)
        )

        elapsed = time.time() - start_time
        size_kb = os.path.getsize(output_path) / 1024
        print(f"✅ Audio Generated in {elapsed:.1f}s ({size_kb:.0f} KB): {output_path}", flush=True)
        return output_path

    except Exception as e:
        print(f"❌ Edge TTS Failed: {e}", flush=True)
        return None