Piper is a fast, local neural text-to-speech engine that uses VITS models exported to ONNX for inference. The repository provides a comprehensive application stack: a Python package (`piper-tts`) with a clean API for voice synthesis, a CLI tool, an HTTP server with a web interface (including viseme animation), and a C/C++ shared library (`libpiper`) for embedding in other applications. It also includes a full training pipeline using PyTorch Lightning, allowing users to fine-tune or train new voices from scratch, with support for multi-speaker models, custom phonemes, and VAD silence trimming. The architecture is modular: phonemization is handled by language-specific backends (espeak-ng, Chinese pinyin/g2pW, Hebrew Nakdimon, Japanese OpenJTalk, Arabic tashkeel), and the core inference engine uses ONNX Runtime for efficient, cross-platform execution. The codebase is well-structured with clear separation of concerns, extensive testing, and CI/CD pipelines that build wheels for all major platforms. Although primarily an application, the training code (VITS model implementations, dataset caching, monotonic alignment Cython) exhibits data/research characteristics, making it a hybrid project. The repository is actively maintained, with a responsive issue tracker and regular updates, though the last commit date appears anomalous.

This repository implements a Linux-based voice satellite for Home Assistant, acting as a local voice assistant that connects to Home Assistant via the ESPHome protocol (aioesphomeapi). The application runs as a Python asyncio service, handling wake word detection using TensorFlow Lite models (MicroWakeWord and OpenWakeWord), audio input via the soundcard library, audio output via libmpv, and optional WebRTC noise suppression/gain. It also provides a WebSocket-based Peripheral API (port 6055) that allows external hardware controllers (LED rings, buttons, HAT boards) to register entities and receive real-time state events. The system is deployed via Docker with docker-compose and includes example peripheral implementations for devices like the ReSpeaker HATs, Jabra Speak 410, and Satellite1 HAT. The architecture is modular, with separate modules for the ESPHome API server (api_server.py), core voice pipeline protocol (satellite.py), entity management (entity.py), wake word loading (wake_word.py), audio playback (player/libmpv.py), and peripheral integration (peripheral_api.py). The application is configured via environment variables and command-line arguments, with preferences persisted to a JSON file. The codebase includes unit tests for most modules, CI/CD workflows, and a Docker build pipeline. External dependencies include aioesphomeapi, websockets, soundcard, python-mpv, webrtc-noise-gain, pymicro-wakeword, pyopen-wakeword, and zeroconf. Key data flows: Microphone audio is captured, optionally processed by WebRTC for noise suppression and gain, then fed into wake word detection. Upon wake word detection, the voice pipeline sends audio to Home Assistant for speech recognition and intent handling. Responses (TTS, media) are received and played via MPV. The Peripheral API allows external hardware to send button events and receive state changes (e.g., listening, thinking, muted). The system supports multiple wake words, stop word detection, timers, and media playback with playlist support.