A high-performance, low-latency, real-time voice conversion system powered by ONNX Runtime and Retrieval-based Voice Conversion (RVC). This application enables real-time voice conversion from a microphone/browser source to a designated target character model with minimal processing latency.

✨ Key Features

🚀 WebSocket Audio Pipeline: Streaming audio transfer using binary WebSocket connections (raw PCM float32) for minimal overhead.
⚡ Multi-Backend ONNX Acceleration: Supports execution providers including NVIDIA CUDA, AMD/Intel DirectML, and fallback CPU.
🎼 High-Fidelity DSP Pipeline:
- Low-Cut Filter: Active 1st order Butterworth high-pass filter at 80Hz to eliminate AC hum and rumble.
- Noise Gate: Threshold-based noise suppression to bypass inference during silence (saving CPU/GPU cycles).
- Gain Controls: Independent input/output digital gain staging.
🧠 Advanced Pitch Extraction: Optimized 16kHz pitch prediction using the RMVPE (Retrieval-based Minimum Vocal Pitch Estimation) model.
🌐 Dual Routing Architecture: Supports routing audio via the web browser (Web Audio API) or directly through the server's local audio hardware (using sounddevice).

🛠️ System Architecture

graph TD
    A[Microphone / Web Browser] -->|Web Audio API| B(WebSocket Connection)
    B -->|Raw Float32 PCM Chunk| C[server.py Backend]
    C -->|1. High-Pass Filter 80Hz| D[DSP Stage]
    D -->|2. Gain & Noise Gate| D
    D -->|3. Resample to 16kHz| E[Hubert/ContentVec ONNX]
    D -->|4. Pitch Estimation RMVPE| F[Pitch Predictor]
    E --> G[RVC ONNX Model Inference]
    F --> G
    G -->|Target Audio Chunks| H(WebSocket Connection)
    H -->|Play audio| I[Browser Speakers / Audio Device]

📁 Repository Structure

server.py — The main WebSocket backend and static HTTP server managing connection loops, audio resampling, and model execution.
start.bat — Windows launcher batch file that automatically resolves the Python virtual environment and executes the server.
requirements.txt — Python dependencies list.
frontend/ — Contains client-side Web UI files:
- frontend/index.html — Control interface layout.
- frontend/app.js — WebSocket communication and client-side audio rendering.
- frontend/styles.css — Custom dashboard styling.
lib/ — core package containing inference models and prediction tools.
weights/ — Directory for voice model weights. Place your custom .onnx and .pth model sub-directories here.
pretrained/ — Directory containing base pre-trained models such as vec-768-layer-12.onnx or vec-256-layer-12.onnx.

🚀 Getting Started

📋 Prerequisites

Python 3.10+ (Recommended)
FFmpeg installed and added to the system PATH (Required for audio processing utilities).
(Optional) NVIDIA CUDA Toolkit (v11.x/12.x) and cuDNN for GPU execution acceleration.

📦 Installation

Clone this repository to your local directory.
Initialize and activate a virtual environment (optional but recommended):
```
python -m venv venv
.\venv\Scripts\activate
```
Install the required dependencies:
```
pip install -r requirements.txt
```
Place your ContentVec base model (vec-768-layer-12.onnx or vec-256-layer-12.onnx) inside the pretrained/ directory.
Place your character models in weights/ in structured folders (e.g., weights/HuTao/ containing HuTao.onnx and HuTao.pth).

🏃 Running the Server

Option A: Quick Launch (Windows)

Simply double-click the start.bat file. It will automatically detect Python, set up the directory paths, and launch the service.

Option B: Manual CLI execution

Execute the server using your terminal:

python server.py --host 127.0.0.1 --port 8765 --http_port 8000 --device cuda

⚙️ Command-Line Arguments

Argument	Description	Default
`--host`	The address the WebSocket server binds to.	`127.0.0.1`
`--port`	WebSocket communication port.	`8765`
`--http_port`	Port serving the static frontend Web UI.	`8000`
`--device`	The ONNX Runtime execution device (`cpu`, `cuda`, `dml`).	`cuda`
`--model`	Target folder name in `weights/` to load directly upon startup.	`None`

Once the server begins execution, it will spin up the local server, and your Web UI should open automatically at http://localhost:8000.

🔊 Audio DSP Details

To achieve low latency without output artifacts, the audio processing utilizes:

Sliding Window Context Buffer: Keeps a short historical buffer of the audio to feed the model the required context frames while minimizing output audio delay.
Convolution Padding Fadeout: 120ms of trailing silent padding is temporarily appended to input segments to avoid edge-fading anomalies inherent to RVC convolutional steps.
Linear Resampling: Low-overhead linear interpolation for quick sample rate adaptation.