local-voice-ai

Why we included this project

Local Voice Agent bundles streaming speech recognition, a local language model, and speech synthesis into one supervised application, so you get a complete spoken dialogue loop rather than a single component. A setup launcher inspects your hardware and memory budget and picks a matching model stack, with profiles for plain CPU, NVIDIA GPUs, Apple Silicon, and a Jetson Orin Nano that can hold real-time conversations on a small edge device. The interface lives in the browser, and a remote-client mode lets a lightweight laptop talk to a voice server running elsewhere on the network. Because the whole stack runs in one container, the assistant stays fully offline, and the launcher handles the fiddly parts of model selection and service startup.