Developer releases offline keyword spotting engine supporting Chinese and English
A developer has built an open-source, portable keyword spotting engine called wakeword_engine.py, designed to enable offline voice commands in software projects. The single-file tool uses a causal TCN model on mel spectrograms and runs via ONNX Runtime, with a model size of roughly 135KB capable of detecting 2–10 keywords in under 5 milliseconds. Development began with Chinese — a tonal, single-syllable-heavy language — which made the architecture language-agnostic, allowing English support to follow with minimal additional work. The engine features a five-layer noise filtering system and currently supports Python, web, and Android platforms, with ESP32 support in progress. The project is available on GitHub under an MIT license and includes early integration support for Home Assistant via the Wyoming protocol.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in