[DEEP-DIVE] Scaffold component — #830

scaffold/deep-dive/tts/tts_pipeline.py

@@ -0,0 +1,99 @@
+#!/usr/bin/env python3
+"""
+TTS Pipeline for Deep Dive
+Converts briefing text to audio via Piper (local) or API
+"""
+
+import subprocess
+import tempfile
+import os
+from pathlib import Path
+from typing import Optional
+
+# Piper configuration
+PIPER_MODEL = "en_US-lessac-medium"  # Good quality, reasonable speed
+PIPER_MODEL_URL = f"https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/lessac/medium/{PIPER_MODEL}.onnx"
+PIVER_CONFIG_URL = f"https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/lessac/medium/{PIPER_MODEL}.onnx.json"
+
+class TTSGenerator:
+    def __init__(self, output_dir: str = "./audio_output"):
+        self.output_dir = Path(output_dir)
+        self.output_dir.mkdir(exist_ok=True)
+        self.model_path = self._ensure_model()
+    
+    def _ensure_model(self) -> Path:
+        """Download Piper model if not present."""
+        model_dir = Path("./piper_models")
+        model_dir.mkdir(exist_ok=True)
+        
+        model_file = model_dir / f"{PIPER_MODEL}.onnx"
+        config_file = model_dir / f"{PIPER_MODEL}.onnx.json"
+        
+        if not model_file.exists():
+            print(f"Downloading Piper model...")
+            subprocess.run(["curl", "-L", "-o", str(model_file), PIPER_MODEL_URL], check=True)
+            subprocess.run(["curl", "-L", "-o", str(config_file), PIVER_CONFIG_URL], check=True)
+        
+        return model_file
+    
+    def generate_audio(self, text: str, output_name: str = None) -> Path:
+        """Generate audio from text using Piper."""
+        output_name = output_name or f"briefing_{datetime.now().strftime('%Y%m%d')}"
+        output_wav = self.output_dir / f"{output_name}.wav"
+        
+        # Piper command
+        cmd = [
+            "piper",
+            "--model", str(self.model_path),
+            "--output_file", str(output_wav)
+        ]
+        
+        # Run Piper
+        process = subprocess.Popen(
+            cmd,
+            stdin=subprocess.PIPE,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
+            text=True
+        )
+        stdout, stderr = process.communicate(input=text)
+        
+        if process.returncode != 0:
+            raise RuntimeError(f"Piper failed: {stderr}")
+        
+        return output_wav
+    
+    def convert_to_opus(self, wav_path: Path) -> Path:
+        """Convert WAV to Opus for Telegram (smaller, better quality)."""
+        opus_path = wav_path.with_suffix(".opus")
+        
+        cmd = [
+            "ffmpeg", "-y",
+            "-i", str(wav_path),
+            "-c:a", "libopus",
+            "-b:a", "24k",  # Good quality for speech
+            str(opus_path)
+        ]
+        
+        subprocess.run(cmd, check=True, capture_output=True)
+        return opus_path
+
+def generate_briefing_audio(text: str, output_dir: str = "./audio_output") -> Path:
+    """Convenience function: text → opus for Telegram."""
+    tts = TTSGenerator(output_dir)
+    wav = tts.generate_audio(text)
+    opus = tts.convert_to_opus(wav)
+    
+    # Clean up WAV
+    wav.unlink()
+    
+    return opus
+
+if __name__ == "__main__":
+    # Test with sample text
+    sample = "This is a test of the Deep Dive briefing system. Piper TTS is running locally."
+    try:
+        result = generate_briefing_audio(sample)
+        print(f"Generated: {result}")
+    except Exception as e:
+        print(f"TTS failed (expected if Piper not installed): {e}")