This comprehensive guide walks through building a completely offline voice assistant on Android using open-source tools. The tutorial covers setting up Whisper for speech recognition, MLC LLM for on-device reasoning, and system TTS for voice output, all running locally through Termux. It progressively adds advanced features
Table of contents
Table of ContentsSystem OverviewRequirementsStep 1: Test Microphone and Audio Playback on AndroidStep 2: Install and Run Whisper for ASRStep 3: Install a Local LLM with MLCStep 4: Local Text-to-Speech (TTS)Step 5: The Core Voice LoopStep 6: Tool Calling (Make It Act)Step 7: Memory and PersonalizationStep 8: Retrieval-Augmented Generation (RAG)Step 9: Multi-Step Agentic WorkflowConclusion and Next StepsSort: