# Text-to-Speech Pipeline with Kokoro TTS A Python script that converts text into natural-sounding speech using the Kokoro TTS engine. The script processes a transcript file, generates speech segments, and merges them into a single audio file. ## Features: - Reads text from a transcript file - Generates speech segments with customizable voice an…

readme.md

Text-to-Speech Pipeline with Kokoro TTS

A Python script that converts text into natural-sounding speech using the Kokoro TTS engine. The script processes a transcript file, generates speech segments, and merges them into a single audio file.

Features:

• Reads text from a transcript file
• Generates speech segments with customizable voice and speed settings
• Saves individual audio segments and their corresponding text
• Merges all audio segments into a single WAV file using FFmpeg
• Organizes output in timestamped directories

Dependencies:

• kokoro (TTS engine)
• soundfile
• ffmpeg (system requirement)

Usage:

1. Place your text in transcript.txt in the same directory as the script
2. Run the script: python tts-kokoro.py
3. Find the output in a timestamped directory (e.g., output_20240321_143022)

The script uses American English voice 'af_bella' with slightly reduced speed (0.9) and splits text on newlines.

transcript.txt

Hello and welcome to Lesson 1.2.2 of our Machine Learning Foundations for Software Engineers course. In the previous lesson, we covered the essential building blocks of machine learning—features, labels, models, and the distinction between training data and new data. Today, we'll dive deeper into the three main types of machine learning: supervised learning, unsupervised learning, and reinforcement learning.

tts-kokoro.py

import os
import soundfile as sf
from kokoro import KPipeline
from datetime import datetime
import subprocess

def read_transcript(file_path):
    """Read text from transcript file."""
    try:
        with open(file_path, 'r', encoding='utf-8') as f:
            return f.read()
    except FileNotFoundError:
        print(f"Error: Could not find {file_path}")
        raise
    except Exception as e:
        print(f"Error reading file: {e}")
        raise

def generate_speech(text, output_dir):
    """Generate speech segments from text."""
    # Initialize pipeline
    pipeline = KPipeline(lang_code='a')  # 'a' for American English

    # Generate audio
    generator = pipeline(
        text, 
        voice='af_bella',
        speed=0.9, 
        split_pattern=r'\n+'
    )

    # Process and save audio files
    for i, (gs, ps, audio) in enumerate(generator):
        print(f"\nProcessing segment {i}")
        print(f"Text: {gs}")  # graphemes/text
        print(f"Phonemes: {ps}")
        
        # Save audio file
        output_file = os.path.join(output_dir, f'sentence_{i:03d}.wav')
        sf.write(output_file, audio, 24000)
        
        # Save text content
        text_file = os.path.join(output_dir, f'sentence_{i:03d}.txt')
        with open(text_file, 'w', encoding='utf-8') as f:
            f.write(gs)
            
        print(f"Saved audio file: {output_file}")
        print(f"Saved text file: {text_file}")

def merge_audio_files(output_dir):
    """Merge all audio segments using ffmpeg."""
    # Create a file list for ffmpeg
    file_list_path = os.path.join(output_dir, 'files.txt')
    
    # Get all wav files and sort them
    wav_files = sorted([f for f in os.listdir(output_dir) 
                       if f.startswith('sentence_') and f.endswith('.wav')])
    
    # Create the files.txt for ffmpeg
    with open(file_list_path, 'w') as f:
        for wav_file in wav_files:
            f.write(f"file '{wav_file}'\n")
    
    merged_output = f'merged.wav'
    
    # Merge all files using ffmpeg
    ffmpeg_cmd = [
        'ffmpeg',
        '-f', 'concat',
        '-safe', '0',
        '-i', 'files.txt',
        '-c', 'copy',
        merged_output
    ]
    
    print("\nMerging audio files...")
    subprocess.run(ffmpeg_cmd, cwd=output_dir)
    print(f"Merged audio saved as: {output_dir}/{merged_output}")

def main():
    # Create output directory with timestamp
    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
    output_dir = f'output_{timestamp}'
    os.makedirs(output_dir, exist_ok=True)
    
    # Get script directory and transcript path
    script_dir = os.path.dirname(os.path.abspath(__file__))
    transcript_path = os.path.join(script_dir, 'transcript.txt')
    
    try:
        # Read transcript
        print(f"Reading transcript from {transcript_path}...")
        text = read_transcript(transcript_path)
        
        # Generate speech segments
        print(f"\nGenerating audio files in {output_dir}...")
        generate_speech(text, output_dir)
        
        # Merge audio files
        merge_audio_files(output_dir)
        
    except Exception as e:
        print(f"An error occurred: {str(e)}")
        return 1
    
    return 0

if __name__ == "__main__":
    exit(main())