speak_ros

A plugin-based text-to-speech system for ROS 2

Architecture

speak_ros provides a flexible TTS framework built on:

• Action-based interface - Asynchronous speech synthesis with progress feedback and cancellation support
• Plugin architecture - Switchable TTS engines via pluginlib without code changes
• Streaming support - Chunked audio synthesis and playback for low-latency response
• PortAudio integration - Cross-platform audio playback

The core package loads TTS plugins dynamically and manages the synthesis pipeline, from text input through audio generation to speaker output.

Packages

Package	Description
speak_ros	Core TTS action server implementation
speak_ros_interfaces	Custom message, service, and action definitions
speak_ros_aivis_plugin	Aivis Speech Engine TTS plugin (Japanese)
speak_ros_open_jtalk_plugin	Open JTalk TTS plugin (Japanese)
speak_ros_openai_tts_plugin	OpenAI TTS API plugin
speak_ros_voicevox_plugin	VOICEVOX TTS plugin (Japanese)

Available Plugins

• Aivis Speech Engine - VOICEVOX-compatible Japanese synthesis with emotion control
• Open JTalk - Offline Japanese speech synthesis
• OpenAI TTS - High-quality cloud-based synthesis via OpenAI API
• VOICEVOX - Japanese synthesis with diverse character voices

Building

Prerequisites

• ROS 2 (Humble or later)
• PortAudio development libraries

Install Dependencies

cd <your_ros2_workspace>
rosdep install --from-paths src --ignore-src -r -y

Build

colcon build --packages-up-to speak_ros speak_ros_aivis_plugin speak_ros_open_jtalk_plugin speak_ros_openai_tts_plugin speak_ros_voicevox_plugin
source install/setup.bash

Usage

Running the TTS Server

Start the TTS action server with the default plugin (Open JTalk):

ros2 run speak_ros speak_ros_node

To use a different plugin, specify the plugin parameter:

# OpenAI TTS
ros2 run speak_ros speak_ros_node --ros-args -p plugin:=speak_ros_openai_tts_plugin::OpenAITTSPlugin

# Aivis Speech Engine
ros2 run speak_ros speak_ros_node --ros-args -p plugin:=aivis_plugin::AivisPlugin

# VOICEVOX
ros2 run speak_ros speak_ros_node --ros-args -p plugin:=voicevox_plugin::VoiceVoxPlugin

Testing with the Test Client

Use the provided test client to send text for synthesis:

# In another terminal
ros2 run speak_ros test_client

The test client sends a sample text to the /speak action server and displays feedback during synthesis and playback.

Listing Available Plugins

To see all installed TTS plugins:

ros2 run speak_ros list_plugins

Action Interface

The /speak action accepts text and optional parameter overrides:

ros2 action send_goal /speak speak_ros_interfaces/action/Speak "{text: 'Hello world'}"

License

Apache 2.0