Kokoro Text to Speech MCP Server

import soundfile as sf from kokoro_onnx import Kokoro import re import os import subprocess from dotenv import load_dotenv load_dotenv() class KokoroTTSService: def __init__(self, model_path="kokoro-v1.0.onnx", voices_path="voices-v1.0.bin"): """Initialize the Kokoro TTS service.""" try: self.kokoro = Kokoro(model_path, voices_path) self.available = True except Exception as e: print(f"Error initializing Kokoro TTS: {e}") self.available = False def get_voices(self): """Return a list of available voices.""" if self.available: return self.kokoro.get_voices() else: # Return default voices if Kokoro is not available return ["af_heart", "en_us_male", "en_us_female"] def generate_audio(self, text, voice="af_heart", speed=1.0, lang="en-us", output_file="audio.mp3", output_dir="mp3"): """Generate audio from text using Kokoro TTS.""" # Clean up the text (remove markdown links) text = self._remove_markdown_links(text) # Create output directory if it doesn't exist os.makedirs(output_dir, exist_ok=True) # Ensure output filename has the correct extension if not output_file.lower().endswith('.mp3'): output_file = f"{os.path.splitext(output_file)[0]}.mp3" if not self.available: # Use fallback TTS method return self._generate_with_fallback(text, output_file, output_dir) try: # Create full file paths base_filename = os.path.splitext(output_file)[0] wav_file = os.path.join(output_dir, f"{base_filename}.wav") mp3_file = os.path.join(output_dir, output_file) # Generate audio samples, sample_rate = self.kokoro.create( text, voice=voice, speed=speed, lang=lang ) # Write audio to WAV file first sf.write(wav_file, samples, sample_rate) # Convert WAV to MP3 success = self._convert_wav_to_mp3(wav_file, mp3_file) # Clean up the WAV file if success and os.path.exists(wav_file): os.remove(wav_file) return { "success": success, "mp3_file": mp3_file if success else None } except Exception as e: print(f"Error generating audio with Kokoro: {e}") return self._generate_with_fallback(text, output_file, output_dir) def _generate_with_fallback(self, text, output_file, output_dir): """Use system TTS as a fallback method.""" try: # Create full file paths base_filename = os.path.splitext(output_file)[0] wav_file = os.path.join(output_dir, f"{base_filename}.wav") mp3_file = os.path.join(output_dir, output_file) # Use macOS 'say' command or other system TTS cmd = ['say', '-o', wav_file, text] subprocess.run(cmd, check=True, capture_output=True, text=True) # Convert to MP3 success = self._convert_wav_to_mp3(wav_file, mp3_file) # Clean up WAV file if success and os.path.exists(wav_file): os.remove(wav_file) return { "success": success, "mp3_file": mp3_file if success else None } except Exception as e: print(f"Error with fallback TTS: {e}") return { "success": False, "mp3_file": None } def _remove_markdown_links(self, text): """Remove markdown links from text.""" # Remove inline links like [text](url) text = re.sub(r'\[([^\]]+)\]\([^)]+\)', r'\1', text) # Remove reference-style links like [text][ref] text = re.sub(r'\[([^\]]+)\]\[[^\]]*\]', r'\1', text) # Remove reference link definitions like [ref]: url text = re.sub(r'^\s*\[[^\]]+\]:\s*.*$', '', text, flags=re.MULTILINE) return text def _convert_wav_to_mp3(self, wav_file, mp3_file): """Convert WAV file to MP3 using ffmpeg.""" try: cmd = ['ffmpeg', '-y', '-i', wav_file, '-codec:a', 'libmp3lame', '-qscale:a', '2', mp3_file] subprocess.run(cmd, check=True, capture_output=True, text=True) return True except Exception as e: print(f"Error converting to MP3: {e}") return False