PyKokoro Documentation

PyKokoro is a Python library for text-to-speech synthesis using the Kokoro TTS model. It provides high-quality, natural-sounding speech generation with support for multiple languages, voices, and advanced features like pause control and text splitting.

Contents:

Features

  • High-Quality TTS: Natural-sounding speech synthesis using the Kokoro model

  • Multiple Languages: Support for English (US/GB), Spanish, French, German, Italian, Portuguese, Hindi, Japanese, Korean, and Chinese

  • Voice Selection: Choose from 54 voices (v1.0) or 103 voices (v1.1-zh) with various styles and accents

  • Voice Blending: Create custom voice styles by blending multiple voices

  • Text Normalization: Automatic say-as support for numbers, dates, phone numbers, and more using SSMD markup

  • Pause Control: Add precise pauses using SSMD break syntax: …c, …s, …p, …500ms

  • Smart Text Splitting: Automatically split long text at sentence, paragraph, or clause boundaries

  • Language-Aware spaCy Models: Auto-resolve spaCy package names from language + size

  • Phoneme-Based Generation: Generate speech directly from phonemes for precise control

  • GPU Acceleration: Optional GPU support for faster generation

  • Flexible Audio Processing: Trim silence, adjust speed, and more

  • Composable Pipeline Stages: Swap doc parsing, splitting, g2p, and audio stages

Quick Example

from pykokoro import GenerationConfig, KokoroPipeline, PipelineConfig

config = PipelineConfig(
    voice="af_bella",
    generation=GenerationConfig(speed=1.0),
)
pipe = KokoroPipeline(config)
result = pipe.run("Hello, world! This is a test.")

# Save to file
import soundfile as sf
sf.write("output.wav", result.audio, result.sample_rate)

Installation

Install via pip:

pip install pykokoro

Or with GPU support:

pip install pykokoro[gpu]

Requirements

  • Python 3.9 or higher

  • NumPy

  • ONNX Runtime

  • espeak-ng (for phonemization)

  • Optional: GPU with CUDA/ROCm for acceleration

Getting Help

License

PyKokoro is released under the MIT License. The Kokoro model itself is subject to its own license terms.

Indices and tables