OSS Alternative - Discover Top Open Source Alternatives to Popular Software

yl4579/StyleTTS2

StyleTTS 2 is a text-to-speech model that achieves human-level speech synthesis by leveraging style diffusion and adversarial training with large speech language models.

Core Features

Achieves human-level TTS synthesis on single and multi-speaker datasets.

Utilizes style diffusion models to generate suitable styles without requiring reference speech.

Employs large pre-trained Speech Language Models (SLMs) as discriminators for improved naturalness.

Enables efficient latent diffusion and diverse speech synthesis.

Supports zero-shot speaker adaptation, outperforming previous publicly available models.

Quick Start

pip install -r requirements.txt

Detailed Introduction

StyleTTS 2 is an advanced text-to-speech (TTS) model pushing the boundaries of synthetic speech naturalness. It innovatively combines style diffusion models to generate appropriate speech styles from text alone, eliminating the need for reference audio. Furthermore, it integrates large pre-trained Speech Language Models (SLMs) like WavLM into an adversarial training framework, significantly enhancing the naturalness of the synthesized speech. This approach has enabled StyleTTS 2 to achieve human-level TTS quality on both single and multi-speaker datasets, and demonstrates superior performance in zero-shot speaker adaptation, marking a significant leap in TTS technology.