wenet-e2e/wekws

Production First and Production Ready End-to-End Keyword Spotting Toolkit

What it solves

WeKws is a production-ready toolkit for small-footprint keyword spotting (KWS), specifically designed for wake-up word (WuW) detection in Internet of Things (IoT) devices. It addresses the need for low-power, low-latency, and computationally efficient models that can run locally and persistently on embedded hardware.

How it works

The toolkit provides an end-to-end framework for developing and training models that detect predefined keywords in a streaming manner. It supports various hardware platforms including x86, Android, Raspberry Pi, and web browsers, ensuring that the detection system can actually be deployed to the IoT devices it is intended for.

Who it’s for

Developers building voice-controlled IoT devices that require hands-free interaction via wake-up words.

Highlights

  • Production-Ready: Focused on efficiency and deployment on resource-constrained hardware.
  • Versatile Detection: Supports single, multiple, customizable, and personalized wake-up words (combining detection with voiceprint).
  • Broad Hardware Support: Compatible with x86, Android, Raspberry Pi, and web browsers.
  • Dataset Support: Compatible with open-source datasets like Hey Snips, Google Speech Command, and Hi Miya.

Related

  • Project
    Picovoice/porcupine

    A lightweight and highly accurate wake word engine for building always-listening voice-enabled applications across IoT, mobile, and web platforms.

  • Project
    jxlarrea/voice-satellite-card-integration

    A Home Assistant integration that transforms any web browser or Android tablet into a private, hands-free voice assistant satellite with on-device wake word detection.

  • Project
    wenet-e2e/wenet

    A production-oriented end-to-end speech recognition toolkit that provides full-stack solutions for streaming and non-streaming speech-to-text.

  • Project
    espressif/esp-sr

    ESP-SR is Espressif's official speech-recognition framework for their ESP32 microcontroller line. It packs wake-word detection ("Hi ESP", "Alexa", "你好小智" etc.), offline command recognition (up to 300 phrases in Chinese/English), audio front-end cleaning (echo cancellation, noise suppression, direction-of-arrival), voice-activity detection, and text-to-speech — all designed to run locally on small, low-power embedded chips.

  • Project
    kyutai-labs/pocket-tts

    Pocket TTS is a 100 M‑parameter, CPU‑only text‑to‑speech library that streams audio with ~200 ms latency, supports multiple languages and voice cloning, and offers a Python API, CLI, and local HTTP server. It can be installed via pip (or uv) and runs faster than real‑time on modest hardware, with community‑trained models and several alternative runtimes for browsers and embedded devices.