archive/app_deploy.py

"""
SpeechT5 Armenian TTS - Production Deployment
============================================

Production-ready version for HuggingFace Spaces with robust error handling.
"""

import gradio as gr
import numpy as np
import logging
import time
import os
import sys
from typing import Tuple, Optional, Union

# Setup logging first
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(levelname)s - %(message)s'
)
logger = logging.getLogger(__name__)

# Global pipeline variable
pipeline = None

def safe_import():
    """Safely import the TTS pipeline with fallbacks."""
    global pipeline
    
    try:
        # Add src to path
        current_dir = os.path.dirname(os.path.abspath(__file__))
        src_path = os.path.join(current_dir, 'src')
        if src_path not in sys.path:
            sys.path.insert(0, src_path)
        
        # Import pipeline
        from src.pipeline import TTSPipeline
        
        logger.info("Initializing TTS Pipeline...")
        pipeline = TTSPipeline(
            model_checkpoint="Edmon02/TTS_NB_2",
            max_chunk_length=200,
            crossfade_duration=0.1,
            use_mixed_precision=True
        )
        
        # Optimize for production
        pipeline.optimize_for_production()
        logger.info("TTS Pipeline ready")
        return True
        
    except Exception as e:
        logger.error(f"Failed to initialize pipeline: {e}")
        logger.info("Creating fallback pipeline for testing")
        
        # Create a simple fallback
        class FallbackPipeline:
            def synthesize(self, text, **kwargs):
                # Generate simple tone as placeholder
                duration = min(len(text) * 0.08, 3.0)
                sample_rate = 16000
                samples = int(duration * sample_rate)
                t = np.linspace(0, duration, samples)
                # Create a simple beep
                audio = np.sin(2 * np.pi * 440 * t) * 0.3
                return sample_rate, (audio * 32767).astype(np.int16)
        
        pipeline = FallbackPipeline()
        return False


def generate_audio(text: str) -> Tuple[int, np.ndarray]:
    """
    Generate audio from Armenian text.
    
    Args:
        text: Armenian text to synthesize
        
    Returns:
        Tuple of (sample_rate, audio_data)
    """
    if not text or not text.strip():
        logger.warning("Empty text provided")
        # Return silence
        return 16000, np.zeros(1000, dtype=np.int16)
    
    if pipeline is None:
        logger.error("Pipeline not available")
        return 16000, np.zeros(1000, dtype=np.int16)
    
    try:
        logger.info(f"Processing: {text[:50]}...")
        start_time = time.time()
        
        # Synthesize with basic parameters
        sample_rate, audio = pipeline.synthesize(
            text=text,
            speaker="BDL",
            enable_chunking=True,
            apply_audio_processing=True
        )
        
        duration = time.time() - start_time
        logger.info(f"Generated {len(audio)} samples in {duration:.2f}s")
        
        return sample_rate, audio
        
    except Exception as e:
        logger.error(f"Synthesis error: {e}")
        # Return silence on error
        return 16000, np.zeros(1000, dtype=np.int16)


# Initialize the pipeline
logger.info("Starting TTS application...")
initialization_success = safe_import()

if initialization_success:
    status_message = "✅ TTS System Ready"
else:
    status_message = "⚠️ Running in Test Mode (Limited Functionality)"

# Create the Gradio interface using the simpler gr.Interface
demo = gr.Interface(
    fn=generate_audio,
    inputs=gr.Textbox(
        label="Armenian Text",
        placeholder="Գրեք ձեր տեքստը այստեղ...",
        lines=3,
        max_lines=8
    ),
    outputs=gr.Audio(
        label="Generated Speech",
        type="numpy"
    ),
    title="🎤 Armenian Text-to-Speech",
    description=f"""
    {status_message}
    
    Convert Armenian text to speech using SpeechT5.
    
    **How to use:**
    1. Enter Armenian text in the box below
    2. Click Submit to generate speech
    3. Play the generated audio
    
    **Tips:**
    - Use standard Armenian script
    - Shorter sentences work better
    - Include punctuation for natural pauses
    """,
    examples=[
        "Բարև ձեզ:",
        "Ինչպե՞ս եք:",
        "Շնորհակալություն:",
        "Կեցցե՛ Հայաստանը:",
        "Այսօր լավ օր է:"
    ],
    theme=gr.themes.Default(),
    allow_flagging="never"
)

# Launch the app
if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        share=False
    )