RVC Voicechanger

A premium, real-time AI voice changer desktop application powered by Retrieval-based Voice Conversion (RVC) and Pedalboard DSP.
Experience near-zero latency conversion, automated image scraping for models, and customizable preset controls in a beautiful glassmorphic dark-mode interface.

---

🌟 Key Features

• Real-time AI Voice Conversion: Highly optimized local RVC pipeline utilizing PyTorch, ONNX, and torchfcpe for pitch extraction.
• Integrated DSP Pedalboard: Embedded real-time noise gate, highpass filter, compressor, pitch shifter, and reverb powered by Spotify's pedalboard library.
• Multi-Language Settings (i18n): Native support for English, 日本語 (Japanese), and 简体中文 (Chinese) covering the entire app layout and telemetry panels.
• Light & Dark Themes: Elegant theme toggle with polished, accessible layouts optimized for both dark glassmorphism and high-contrast light modes.
• Automatic Web Image Scraper: Automatically scans .pth and .index model names to scrape, cache, and display high-quality model avatars from DuckDuckGo and Wikipedia.
• One-Click RVC Import: Seamlessly import custom Applio voice models via a drag-and-drop local import interface.
• Compact Model Manager: Star your favorite models or delete custom presets directly from the grid.
• Clean Mode Bypass: Select the Original voice to bypass the DSP chain entirely and route your clean microphone audio directly.
• Low Latency & High Performance: Tailored multi-threaded pipeline with macOS OpenMP conflict mitigation for smooth performance on Apple Silicon.

---

📸 Screenshots

---

🏗️ Architecture & Workflow

The application runs a hybrid desktop architecture: a modern React frontend hosted locally by a Python FastAPI server, all orchestrated within an Electron shell.

graph TD
    A[Electron Shell] <-->|IPC Bridge| B[Next.js Renderer UI]
    A -->|Spawns / Manages| C[FastAPI Python Backend]
    B <-->|REST / WebSockets| C
    C -->|Audio Streaming| D[Sounddevice Input/Output]
    C -->|AI Inference| E[RVC Inference Pipeline]
    C -->|Audio Effects| F[Spotify Pedalboard DSP]
    C -->|Model Manager| G[Web Image Scraper & Disk Cacher]

Loading

---

🛠️ Technology Stack

• Frontend Shell: Electron, JavaScript
• User Interface: React, Next.js, Tailwind CSS, Lucide Icons, Shadcn UI
• Backend Server: FastAPI, Uvicorn, WebSockets
• Audio Engine: PySoundDevice, Spotify Pedalboard, PyTorch, FAISS, Librosa

---

🚀 Getting Started

Prerequisites

• Node.js: v18.x or later (pnpm recommended)
• Python: v3.10 to v3.12 (with pip and venv)

Installation

1. Clone the Repository:

   git clone https://github.com/satiricalguru/RVC-Voicechanger.git
   cd rvc-voicechanger

2. Set Up the Python Virtual Environment:

   python3 -m venv .venv
   source .venv/bin/activate
   pip install -r requirements.txt

3. Install Frontend Dependencies:

   cd ui
   pnpm install
   cd ..

4. Install Electron Shell Dependencies:

   pnpm install

Running the Application

• Development Mode:

  npm run dev

  This launches the FastAPI backend locally, boots the hot-reloaded Next.js client, and opens the Electron desktop window.

• Production Build & Start:

  npm start

Packaging the Application

Compile the app bundle into a standalone executable (e.g., .dmg on macOS, .exe on Windows):

npm run build

🍎 macOS Gatekeeper Bypass

Since these builds are unsigned, macOS Gatekeeper will block first-launch execution and may display a warning that the app is "damaged and cannot be opened".

To bypass this and run the app:

1. Drag RVC Voicechanger.app from the .dmg into your Applications folder.
2. Open your terminal and execute:

   xattr -d com.apple.quarantine /Applications/RVC\ Voicechanger.app

3. Launch the application normally from Finder or your Application list.

---

📂 Folder Structure

.
├── app/                  # FastAPI Backend API
│   ├── backend/          # Models Manager, Engine, and RVC Pipeline
│   └── frontend/         # Static compiled Next.js bundle
├── assets/               # Branding assets & application icons
├── electron/             # Main process & IPC handler definitions
├── ui/                   # Next.js / Tailwind React components
├── models/               # RVC Models folder
│   ├── default/          # Native pre-installed RVC checkpoints
│   ├── custom/           # User imported .pth / .index files
│   └── images/           # Cached scraped web avatars
├── requirements.txt      # Python dependencies
└── package.json          # Node scripts and Electron configuration

---

👥 Contributors & Acknowledgement

Contributors

• Jatin Pandey (@satiricalguru) — Project lead, core developer, and UI designer.

Acknowledgement

• This project builds on top of and acknowledges the outstanding open-source work of IAHispano/Applio which provided the foundation for RVC voice inference structures and configurations.

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.