FatihMakes/Mark-LIII

Newest, latest and most advanced model that is able to control computers and even more. We're improving that model day by day, so follow up to know and see new models.

What it solves

Mark LIII is a cross-platform personal AI assistant designed to provide hands-free, real-time voice interaction and system control. It eliminates the need for subscriptions by leveraging the Gemini Live API, allowing users to control their computers (Windows, macOS, and Linux) via voice commands while maintaining digital autonomy.

How it works

The project uses the Gemini 3.1 Flash Live engine for low-latency audio streaming and conversation. It features a local wake-word detector ("Hey Jarvis") that processes audio on-device to ensure privacy and reduce cloud costs. The system is built with a modular architecture where skills and plugins are self-describing files, allowing new capabilities to be added without modifying the core code. It also implements a local search-based memory system to store and recall personal facts without bloating the system prompt.

Who it’s for

Users looking for a highly customizable, voice-driven AI assistant that can control their OS, manage files, and integrate with various web services without relying on a paid subscription service.

Highlights

  • Hands-Free Control: Local "Hey Jarvis" wake-word detection with auto-sleep functionality.
  • System Integration: Ability to launch apps, adjust system settings (volume, brightness, WiFi), and manage files with an "Undo" feature for reversible actions.
  • Extensible Plugin System: New skills can be added by simply dropping a .py file into the plugins folder.
  • Visual Awareness: Real-time screen capture and webcam vision integrated into the Gemini session.
  • Persistent Memory: A local memory store with a dedicated UI panel for managing stored facts.
  • Low Latency: Powered by Gemini 3.1 Flash Live for faster time-to-first-word responses.
  • Cross-Platform: Native support for Windows, macOS, and Linux with OS-specific dependency handling.

Related

  • Project
    i365dev/free4chat

    An experimental system for temporary, account-free collaboration rooms where humans and independently owned AI agents can exchange real-time media and tasks.

  • Project
    auto-novel/auto-novel

    AutoNovel is an open‑source web service that automatically translates light‑novel text using machine translation and shares the results. It can be run locally with Docker Compose, though the authors warn that personal deployment isn’t officially supported.

  • Project
    Ricky-7-Yan/intelligent-audit-system

    An enterprise AI Agent workbench for audit delivery that automates evidence retrieval, control testing, and remediation within a traceable and evaluatable workflow.

  • Project
    proxysoul/Empryo

    Empryo is a graph-powered AI coding agent that maps entire codebases to understand dependencies and blast radius, enabling precise AST-level edits with lower token costs.