Skip to content

README.md

Latest commit

 

History

History
97 lines (71 loc) · 2.72 KB

README.md

File metadata and controls

97 lines (71 loc) · 2.72 KB

VoiceMorph – Personalized Voice Changer Platform

VoiceMorph is a modular, AI-powered voice conversion platform that allows users to record and send their own voice, train a model based on a target speaker, and transform new voice inputs into the target voice.

The system routes conversion tasks to external rvc APIs (currently Replicate), prioritizes them based on user subscription tier, and will eventually migrate to an in-house, cost-efficient inference service.

🔧 Tech Stack

  • FastAPI for API server
  • Beanie for MongoDB ODM
  • Docker / Docker Compose for containerization
  • fastapi-mongo-base Python framework (published on PyPI) for scaffolding and development structure
  • Replicate (temporary) for voice conversion backend
  • Modular app structure supporting task-based execution

📁 Project Structure

.
├── app/
│   ├── apps/
│   │   ├── neda/          # voice conversion logic
│   │   └── voice/         # voice-related APIs
│   ├── server/            # App server and configuration
│   ├── utils/             # Utilities: finance logic, media handling, etc.
│   └── Dockerfile
├── docker-compose.yml
├── README.md
└── sample.env

🧠 Core Components

  • apps.voice:
    Handles management of available target voice models.

    • schemas.py: Data structures for request/response validation
    • models.py: MongoDB voice model definitions
    • services.py: Business logic for managing voice models
    • routes.py: REST endpoints for adding and listing models

  • apps.neda:
    Contains the core voice conversion APIs.

    • Accepts voice conversion requests from users
    • Sends tasks to the external RVC backend (Replicate)
    • Selects the appropriate target voice model

  • utils.finance:
    Connects to an external finance microservice

    • Fetches user subscription info
    • Prioritizes tasks based on subscription tier

  • server.worker:
    Manages periodic background checks on tasks

    • Detects and retries incomplete conversions
    • Handles failure recovery in case of webhook errors

🚀 Getting Started

1. Clone the repository

git clone https://github.com/your-org/voicemorph.git
cd voicemorph

2. Configure environment

Copy the example environment file and fill in required values:

cp sample.env .env

3. Run with Docker

docker-compose up --build

The app should be available at http://localhost:8000.

📌 Future Roadmap

  • 🔄 Migrate from Replicate to self-hosted GPU inference
  • 📊 Add user dashboard for voice model and task management

📄 License

MIT License