This project implements a Mandarin AI voice assistant with local LLM, ASR, and TTS. It works fully offline and is suitable for low-cost hardware such as RK3568.
Features
Offline Mandarin speech recognition (Vosk)
Local LLM (Qwen or Llama.cpp)
PaddleSpeech TTS
Automatic silence detection and auto-stop recording
Voice or button command to clear context memory
Robust error handling and modular design
How to use
Before run the code download models for Vosk, Qwen (GGUF), and PaddleSpeech (FastSpeech2/HiFiGAN), put them in the models/ folder.
Offline Mandarin AI Voice Assistant Project
This project implements a Mandarin AI voice assistant with local LLM, ASR, and TTS. It works fully offline and is suitable for low-cost hardware such as RK3568.
Features
How to use
Before run the code download models for Vosk, Qwen (GGUF), and PaddleSpeech (FastSpeech2/HiFiGAN), put them in the models/ folder.
use python edge_server.py to run the code