bailing

Why we included this project

This project turns a normal laptop into a voice assistant without needing a GPU, using a pipeline of speech recognition, voice-activity detection, an LLM for reasoning, and speech synthesis. It handles interruption mid-sentence and remembers your preferences and conversation history, which makes the back-and-forth feel less like a chatbot demo and more like a tool you can actually use. Beyond answering questions, it can act through OpenClaw: ask for the weather, set reminders, open apps on macOS, or trigger a web search by voice. The individual components (FunASR for ASR, silero-vad, DeepSeek for the LLM, and several TTS options) are modular and swappable, so a team or hobbyist can adapt the stack to their own needs. That modularity plus the low hardware bar makes it a practical starting point for experimenting with spoken interfaces.