Stealth-mode AI overlay
A 100% unviewable AI assistant for high-stakes interviews and meetings. Bypasses every screen-capture pipeline.
Version 2.0.0
What it does
Three primitives compose the overlay: stealth rendering, system audio capture, and selective vision. Together they produce live intelligence that interviewers and meeting hosts cannot detect or screen-share.
Built on the Windows Display Affinity API, the overlay is invisible to every screen-capture and screen-share pipeline — OBS, Zoom, Teams, Discord, browser share. Interviewers see only your face and shared desktop while the assistant runs locally, on your screen alone.
WASAPI loopback taps your system audio directly for real-time transcription — every voice in the call, including yours. No virtual cables, no plugins, no audio routing. Install once and the assistant hears the meeting exactly as your speakers do.
Selective region capture sends only the pixels you mark — a code editor, a terminal pane, an open spec — to the model. The rest of your desktop stays private. Visual context arrives without ever exposing the full screen or background tabs.
How It Works
From launch to live intelligence — the entire pipeline runs silently in under 2 seconds.
Start the assistant with a single hotkey. The overlay attaches to your display and immediately becomes unviewable to every screen-capture and screen-share pipeline.
The system silently captures meeting audio and selected screen regions, feeding real-time context to the AI model — no virtual cables, no plugins.
AI-generated suggestions, answers, and talking points appear directly on your screen. Only you can see them. Screen recorders and participants see nothing.
Compatibility: Requires Windows 10 version 2004 or later. Works with Zoom, Teams, Google Meet, Discord, OBS, and all DXGI-based capture tools.
Get Started
Pick the guide that matches where you are. Setup walks you through installation in under two minutes; Usage shows you how to drive the overlay during a live call.
