Skip to main content

Overview

revolution/voicevox-core is a pure PHP package that wraps the VOICEVOX CORE C dynamic library via PHP FFI. VOICEVOX is a high-quality Japanese text-to-speech (TTS) engine. With this package, you can perform speech synthesis directly from PHP.
If you want to understand how PHP FFI works and its constraints first, see PHP FFI.
PHP FFI is often disabled in web server environments such as FPM (ffi.enable=false). This package is intended for local CLI use.

Requirements

  • PHP 8.3+
  • ext-ffi extension enabled
  • VOICEVOX CORE 0.16+

Installation

Library Setup

To use this package, you need the VOICEVOX CORE dynamic library (.so / .dylib), the ONNX Runtime library, and the OpenJTalk dictionary.
1

Download voicevox_core

Download the downloader for your OS and architecture from voicevox_core releases and run it. It creates a voicevox_core directory in your current directory.The created directory structure:
2

Move to a permanent location

3

Create a symbolic link (recommended)

Create a symbolic link so the library is found automatically.
Always use absolute paths with ln -s.
macOS:
If the library cannot be loaded from /usr/local/lib/, set DYLD_FALLBACK_LIBRARY_PATH in your .zshrc or similar:
Linux:

Alternative: Use an environment variable

If you cannot create a symbolic link, set the full path to the library file in the VOICEVOX_CORE_LIB_PATH environment variable:
On macOS, if DYLD_FALLBACK_LIBRARY_PATH does not work, use VOICEVOX_CORE_LIB_PATH instead.

Documentation

Usage

Basic usage and code samples for TTS and speech synthesis.

API Reference

Detailed reference for all classes, methods, and enums.
Last modified on May 17, 2026

Related topics

VOICEVOX Core for PHP