OSS Alternative - Discover Top Open Source Alternatives to Popular Software

TensorSpeech/TensorFlowTTS

TensorFlowTTS is a real-time, state-of-the-art speech synthesis library built on TensorFlow 2, supporting multiple languages and optimized for efficient deployment.

Core Features

Real-time state-of-the-art speech synthesis architectures (Tacotron-2, FastSpeech, Melgan, etc.)

Multi-language support including English, French, Korean, Chinese, and German

Optimized for deployment on mobile/embedded systems via TFLite, quantization, and pruning

Scalable training with Multi-GPU support and mixed precision

Easy adaptation and fine-tuning for new languages or models

Detailed Introduction

TensorFlowTTS is an advanced open-source library providing real-time, state-of-the-art speech synthesis capabilities using TensorFlow 2. It integrates various cutting-edge architectures like Tacotron-2, FastSpeech, and Melgan, enabling high-quality text-to-speech conversion. The project emphasizes performance and deployability, leveraging TensorFlow 2's optimizations such as fake-quantize aware training and pruning to achieve faster-than-real-time inference. It supports multiple languages out-of-the-box and is designed for easy adaptation, making it suitable for a wide range of applications from research to production on resource-constrained devices.