VOICE OVER LORA

Abstract

This project presents a novel voice messaging system over LoRa, designed for use in low-bandwidth, infrastructure-limited environments such as disaster zones and remote areas. Instead of transmitting raw audio, the system employs a pipeline combining speech-to-text (Whisper), on-device summarization via a quantized large language model (Ollama), and text-to-speech synthesis (Piper). This approach significantly reduces data size through semantic compression while preserving message intelligibility. Implemented on Raspberry Pi 4 devices with LoRa transceivers, the system operates fully offline and demonstrates the practical integration of edge AI for intelligent, efficient, and human-friendly voice communication.