lunavox-tts 1.0.4

GPT-SoVITS ONNX Inference Engine & Model Converter

 ___       ___  ___  ________   ________  ___      ___ ________     ___    ___ 
|\  \     |\  \|\  \|\   ___  \|\   __  \|\  \    /  /|\   __  \   |\  \  /  /|
\ \  \    \ \  \\\  \ \  \\ \  \ \  \|\  \ \  \  /  / | \  \|\  \  \ \  \/  / /
 \ \  \    \ \  \\\  \ \  \\ \  \ \   __  \ \  \/  / / \ \  \\\  \  \ \    / / 
  \ \  \____\ \  \\\  \ \  \\ \  \ \  \ \  \ \    / /   \ \  \\\  \  /     \/  
   \ \_______\ \_______\ \__\\ \__\ \__\ \__\ \__/ /     \ \_______\/  /\   \  
    \|_______|\|_______|\|__| \|__|\|__|\|__|\|__|/       \|_______/__/ /\ __\ 
                                                                   |__|/ \|__| 

🔮 LunaVox: GPT-SoVITS Lightweight Inference Engine

Experience near-instantaneous speech synthesis on your CPU

简体中文 | English

---

LunaVox is a lightweight inference engine built on the open-source TTS project GPT-SoVITS. It integrates TTS inference, ONNX model conversion, API server, and other core features, aiming to provide ultimate performance and convenience.

• ✅ Supported Model Version: GPT-SoVITS V2
• ✅ Supported Language: Japanese
• ✅ Supported Python Version: >= 3.9

---

🚀 Performance Advantages

LunaVox optimizes the original model for outstanding CPU performance.

Feature	🔮 LunaVox	Official PyTorch Model	Official ONNX Model
First Inference Latency	1.13s	1.35s	3.57s
Runtime Size	~200MB	~several GB	Similar to LunaVox
Model Size	~230MB	Similar to LunaVox	~750MB

📝 Note: Since GPU inference latency does not significantly improve over CPU for the first packet, we currently only provide a CPU version to ensure the best out-of-the-box experience.

📝 Latency Test Info: All latency data is based on a test set of 100 Japanese sentences (~20 characters each), averaged. Tested on CPU i7-13620H.

---

🏁 QuickStart

⚠️ Important: It is recommended to run LunaVox in Administrator mode to avoid potential performance degradation.

📦 Installation

Install via pip:

pip install lunavox-tts

📝 You may encounter an installation failure when trying to install pyopenjtalk. This is because pyopenjtalk is a library that includes C extensions, and the publisher does not currently provide pre-compiled binary packages ( wheels). For Windows users, this requires installing Visual Studio Build Tools. Specifically, you must select the "Desktop development with C++" workload during the installation process.

⚡️ Quick Tryout

No GPT-SoVITS model yet? No problem! LunaVox includes predefined speaker characters for immediate use without any model files. Run the script below to hear it in action:

python Tutorial/quick_tryout.py

This script will automatically download required dependencies and play a sample audio.

🎤 TTS Best Practices

A simple TTS inference example:

import lunavox_tts as lunavox

# Step 1: Load character voice model
lunavox.load_character(
    character_name='<CHARACTER_NAME>',  # Replace with your character name
    onnx_model_dir=r"<PATH_TO_CHARACTER_ONNX_MODEL_DIR>",  # Folder containing ONNX model
)

# Step 2: Set reference audio (for emotion and intonation cloning)
lunavox.set_reference_audio(
    character_name='<CHARACTER_NAME>',  # Must match loaded character name
    audio_path=r"<PATH_TO_REFERENCE_AUDIO>",  # Path to reference audio
    audio_text="<REFERENCE_AUDIO_TEXT>",  # Corresponding text
)

# Step 3: Run TTS inference and generate audio
lunavox.tts(
    character_name='<CHARACTER_NAME>',  # Must match loaded character
    text="<TEXT_TO_SYNTHESIZE>",  # Text to synthesize
    play=True,  # Play audio directly
    save_path="<OUTPUT_AUDIO_PATH>",  # Output audio file path
)

print("🎉 Audio generation complete!")

---

🔧 Model Conversion

To convert original GPT-SoVITS models for LunaVox, ensure torch is installed:

pip install torch

Use the built-in conversion tool:

Tip: convert_to_onnx currently supports only V2 models.

import lunavox_tts as lunavox

lunavox.convert_to_onnx(
    torch_pth_path=r"<YOUR .PTH MODEL FILE>",  # Replace with your .pth file
    torch_ckpt_path=r"<YOUR .CKPT CHECKPOINT FILE>",  # Replace with your .ckpt file
    output_dir=r"<ONNX MODEL OUTPUT DIRECTORY>"  # Directory to save ONNX model
)

---

🌐 Launch FastAPI Server

LunaVox includes a lightweight FastAPI server:

import lunavox_tts as lunavox

# Start server
lunavox.start_server(
    host="0.0.0.0",  # Host address
    port=8000,  # Port
    workers=1  # Number of workers
)

For request formats and API details, see our API Server Tutorial.

---

⌨️ Launch CMD Client

LunaVox provides a simple command-line client for quick testing and interactive use:

import lunavox_tts as lunavox

# Launch CLI client
lunavox.launch_command_line_client()

---

📝 Roadmap

• 🌐 Language Expansion

  • Add support for Chinese and English.

• 🚀 Model Compatibility

  • Support for V2Proplus, V3, V4, and more.

• 📦 Easy Deployment

  • Release Docker images.
  • Provide out-of-the-box Windows / Linux bundles.

---