Open-source Speechify alternatives

The 2noise project publishes ChatTTS, a generative speech model designed for daily dialogue. It turns text into natural-sounding conversational speech and is aimed especially at dialogue scenarios such as the voice of an LLM-powered assistant.

Supported languages are English and Chinese, with more planned. The repository contains the algorithm infrastructure and simple examples, while end-user products built on top of the model are collected in a community-maintained index called Awesome-ChatTTS. The model is implemented in Python with PyTorch and torchaudio, and the README is translated into several languages. The project highlights conversational speech with control over prosodic details such as laughter and pauses.

The code is released under AGPL-3.0 and runs on your own hardware, typically with a GPU for good speed. The model weights may carry separate terms from the code, so review those terms before any commercial use.

Piper is a neural text-to-speech engine that produces spoken audio on your own computer instead of sending text to a cloud service. It embeds espeak-ng to convert text into phonemes, and a trained voice model then turns those phonemes into speech. The project is aimed at keeping synthesis fast and fully local.

Piper can be used from a command-line interface, through an HTTP web server, from a Python API or as a C/C++ library, so it fits scripts, services and embedded applications alike. The repository links to audio samples, a browser demo, a catalog of voices, and documentation for training new voices and building from source.

Piper1 is released under the GPL-3.0 licence and is written mostly in C++. The Open Home Foundation, which hosts the project, has asked for new maintainers. Its README lists many projects that build on it, among them Home Assistant, the NVDA screen reader and LocalAI, and community bindings exist for other languages such as Java.