On-device wake word detection powered by deep learning
This is a list of interesting papers and projects about TinyML.
Turn an Echo Dot (2nd gen) into a fully local Home Assistant voice satellite, multi-room media player, lux sensor, Bluetooth proxy, and more. No Amazon, no cloud.
OpenClaw voice assistant app for Android - Wake word activation & system assistant integration
Open firmware for the Echo Show 5, 1st and 2nd gen: Alpine Linux instead of Android, no Alexa. A local Home Assistant voice satellite with its own touch screen, on-device wake word, SSH and A/B over-the-air updates.
Dashie for Android — the family dashboard and Home Assistant kiosk app. AGPL-3.0.
wake word engine benchmark framework
A lightweight, wake word detection engine. Train custom, high-accuracy models with minimal effort.
Linux firmware for the Echo Dot (2nd gen): no Fire OS, no Alexa. A local Home Assistant voice satellite with all seven mics, Bluetooth speaker mode, signed A/B updates and SSH.
Your AI assistant, built for the agentic era. Open source and local: talk to it, and it runs your agents, your coding CLIs, your browser and your apps. Windows, macOS, Linux.
Offline keyword spotting & wake word engine. ONNX format, model <130KB, inference <10ms. Runs on Android, ESP32, Linux, Web. No cloud, no data usage.
Create any wake word, deploy on any device and operating system.
Wake word detection with custom phrases without model training
A custom "Hey Donut" wake word for Home Assistant — 62 KB, runs on-device, zero false accepts per hour. Includes the full training pipeline so you can teach your assistant any name. CC0.
GPT powered personal voice assistant, immitates GLaDOS from Portal.
A lightweight, offline-capable voice assistant that listens for a wake phrase, converts speech to text, runs local or API-based intent handling, and replies with natural-sounding TTS. Built for quick local deployment, it supports configurable wake words, microphone I/O, and easy integration with custom actions or LLMs.
Use Porcupine wake word to trigger a voice assistant pipeline in Home-Assistant
Local-first voice assistant with dual-brain architecture: fast conversational LLM + tool-calling agent. Wake-word triggered, half-duplex, all models run on your own GPU.
OVOS voice satellite for HiveMind — full on-device voice stack (microphone, VAD, wake word, STT, TTS)
Local-first personal wake-word training with ONNX export.
To associate your repository with the wake-word topic, visit your repo's landing page and select "manage topics."