Luigi Maselli is a software developer who publishes as grigio. His work includes local voice interfaces and obs-cmd, a command-line client for controlling OBS Studio through its WebSocket interface.[3][4][1]
Start with Offline voice assistant, NO dedicated GPU, less 15Gb RAM →
GitHub 
X Contributions
Keep listening while a local assistant is speaking
Maselli’s local voice-assistant account describes a continuous microphone loop rather than a push-to-talk exchange. Speech detection, pause timing and an utterance classifier help decide whether to keep listening or reply. Interruptions pause or discard speech, and the conversation retains what was already spoken. A separate gate compares proposed tool calls with the user’s last request. These mechanisms address turn-taking and unintended actions; the article describes his implementation, not a guarantee across machines.[4]
Control streaming software from scripts
obs-cmd exposes OBS Studio operations through the current obs-websocket interface. A shell command can change scenes or control recording and streaming, making those actions usable from scripts and external controls. The OBS process still needs WebSocket access enabled and configured; the client provides a terminal interface to that existing service.[1]
Tools, writing & talks
Writing, research & talks
Start here
An implementation account of continuous local speech, interruptions, tool gating and conversation-state management.
Read the wiring and turn-taking sections before attempting setup. You need a local model service, speech components and command-line familiarity; the article’s hardware claim describes the author’s configuration.
Author and builder[4]Original source