ollama-voice-mac

Why we included this project

For anyone who wants a voice assistant that never sends audio or text anywhere, this project is a straightforward way to get one running entirely on macOS. It captures your microphone input, transcribes it locally with a Whisper model, passes the text to Mistral 7b through Ollama, and replies using the Mac's system voice, so nothing leaves the machine. The code builds on the earlier ollama-voice project with Mac-specific fixes, which keeps it easy to follow if you want to edit the prompts in assistant.yaml, swap in a different model, or add another language by dropping in a matching Whisper model. The installation steps are explicit for macOS, so you can go from cloning the repo to talking in a short session.