This repository serves as a comprehensive, evolving survey of controllable speech synthesis, focusing on recent advancements driven by Large Language Models (LLMs) and diffusion models. It targets researchers and practitioners in the field, providing a structured taxonomy of methods, datasets, and evaluation techniques to facilitate understanding and accelerate development in generating speech with fine-grained control over attributes like emotion, timbre, and style.

How It Works

This project is a curated collection of research papers, not a runnable system. It categorizes controllable Text-to-Speech (TTS) methods based on their architectures, control strategies, and feature representations. The core approach involves surveying and organizing advancements, particularly those leveraging LLMs and diffusion models, to offer a clear taxonomy and highlight emerging trends and future research directions.

Quick Start & Requirements

The repository itself does not provide a single installation or run command as it is a curated list of research papers. However, individual papers listed within the repository often include links to their respective code repositories and demos, allowing users to explore and run specific controllable TTS systems. Users should refer to the linked papers for specific setup instructions and dependencies (e.g., Python versions, GPU requirements, specific libraries) for each individual project.

Highlighted Details

• Features a systematic survey of controllable TTS, categorizing models by architecture, control strategies, and feature representations.
• Meticulously lists and categorizes hundreds of recent research papers (2021-2026) with details on zero-shot capability, controllability attributes, acoustic models, vocoders, and links to demos/code.
• Includes detailed tables summarizing relevant open-source datasets and common evaluation metrics (objective and subjective) used in the field.
• Provides video introductions for the associated survey paper in both English and Chinese.

Maintenance & Community

The repository is explicitly described as "evolving" and actively encourages community contributions. Users are invited to open issues for corrections and submit pull requests to add new papers or relevant resources, fostering a collaborative environment for keeping the survey up-to-date.

Licensing & Compatibility

No specific open-source license is mentioned in the provided README content. This lack of explicit licensing information may pose a barrier for commercial use or integration into closed-source projects.

Limitations & Caveats

As a survey and resource collection, this repository does not offer a unified, ready-to-use controllable speech synthesis system. Users must navigate and set up individual projects linked within the repository. The comprehensiveness and accuracy are dependent on ongoing community contributions and the dynamic nature of the research field.