EaseVoice Trainer is a backend project designed to streamline and enhance the training process for voice synthesis and transformation. It is built upon the foundation of GPT-SoVITS, inheriting its core concepts while introducing various improvements to make the system more accessible, elegant, and user-friendly.
Although EaseVoice Trainer takes inspiration from GPT-SoVITS, we chose not to directly fork the original repository due to significant design differences and unique objectives. Our focus is on creating a refined, modular system tailored to specific use cases and improving maintainability.
- User-Friendly Design: Simplified workflows and intuitive configurations make it easier for users to deploy and manage.
- Stability: ensuring consistent and reliable performance in the process of cloing and training.
- Training Observability: offering comprehensive monitoring tools, providing clear insights into progress of cloning and training with performance metrics.
- Clean Architecture: We have split the project into separate frontend and backend repositories for better modularity and maintainability. The portal repository is at EaseVoice Trainer Frontend.
- RESTful API: The backend provides a RESTful API for easy integration with other services and applications.
- Scalability: Built with scalability in mind, making it suitable for both small-scale experiments and large-scale production.
- Integrate Tensorboard: We have integrated Tensorboard for real-time monitoring and visualization of training progress.
As in GPT-SoVITS, you can download Pretrained Models, and then put them into the models directory.
Before using EaseVoice Trainer, ensure you have the following installed: • Python 3.9 or higher • uv.
We leverage uv for robust and efficient project management. To start the server using uv:
uv sync
uv pip install whl/LangSegment-0.3.5-py3-none-any.whl
uv run src/main.py- Build the Docker image
cd scripts
docker build -t megaease/easevoice-trainer .
- Run the Docker Container
docker run -p 8000:8000 megaease/easevoice-trainer
This command ensures that the application running inside the container on port 8000 is accessible locally via http://localhost:8000.
We welcome contributions from the community! Whether it’s fixing bugs, adding new features, or improving documentation, feel free to submit a pull request or open an issue.
EaseVoice Trainer is under the Apache 2.0 license. See the LICENSE file for details.
