LiveKit Kokoro TTS Plugin

A high-performance text-to-speech plugin for LiveKit agents using Kokoro TTS with ultra-low latency streaming implementation for real-time voice synthesis.

Features

• Ultra-Low Latency: ~80ms time-to-first-byte (TTFB) on RTX 4090
• Multiple Voices: Support for multiple voices and voice mixing
• Streaming Support: Real-time audio generation with chunked streaming
• LiveKit Integration: Seamless integration with LiveKit agents framework

Requirements

• LiveKit Agents v1.0 or higher
• Kokoro FastAPI server instance
• NVIDIA GPU (recommended for optimal performance)
• Python 3.8+

Performance

Hardware	Latency	Quality	Use Case
RTX 4090	~80ms TTFB	High	Real-time applications

Installation

1. Clone or download this plugin into your LiveKit-based agents project root directory
2. Set up the Kokoro FastAPI server for model inference
3. Install required dependencies:

   pip install openai httpx

Server Setup

Use the Kokoro FastAPI server for optimized inference:

Repository: remsky/Kokoro-FastAPI

This server provides OpenAI-compatible endpoints with optimized Kokoro TTS inference for ultra-low latency performance.

Usage

Initialize your agent session with the KokoroTTS plugin:

from kokoro_plugin import KokoroTTS

session = AgentSession(
    # ... other configuration
    tts=KokoroTTS(
        base_url="http://localhost:8000",
        api_key="NULL",
        voice="af_heart",
        speed=1.0
    )
)