Open the PrepPilot Overlay

Launch the PrepPilot desktop app and click the Stealth Mode toggle. A transparent, resizable overlay appears on your screen. It floats above every window but is excluded from all screen-capture APIs at the operating system level.

Technical detail: Built with Tauri and Rust, the overlay uses platform-specific window flags (SetWindowDisplayAffinity on Windows, CGWindowSharingType on macOS) to exclude itself from screen capture. This is not CSS transparency or opacity tricks — it is genuine OS-level invisibility.

Join Your Video Call

Open Zoom, Google Meet, Microsoft Teams, Webex, or any other video conferencing platform. The PrepPilot overlay sits on top of your call window. You can reposition it, resize it, or adjust its opacity to match your setup.

Works everywhere: Because PrepPilot is a native desktop application, it is completely platform-agnostic. It works with browser-based calls, desktop apps, and even phone-based video platforms when mirrored to your screen.

Interviewer Speaks — Deepgram Transcribes

When the interviewer asks a question, Deepgram captures system audio (not your microphone) and converts speech to text in real time. The transcription appears in the overlay with sub-300ms latency, supporting 30+ languages with automatic language detection.

Audio pipeline: PrepPilot captures system audio using virtual audio loopback, meaning it hears exactly what you hear. It does not access your microphone, so there is zero risk of feedback or echo. Deepgram's Nova-2 model handles accents, technical jargon, and background noise with industry-leading accuracy.

AI Generates a Direct Answer

The transcribed question is sent to your chosen AI model. You can select from Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5, GPT-5.3, or Gemini. The model generates a structured, interview-ready answer in 1-3 seconds.

Context awareness: PrepPilot maintains conversation context throughout the interview. The AI remembers previous questions and your job description (if uploaded), so answers are tailored to the specific role and build on earlier responses.

Read and Speak Naturally

The AI-generated answer appears in the overlay. Read the key points and deliver them in your own voice and style. The overlay is positioned so you can glance at it naturally while maintaining eye contact with the camera.

Pro tip: Position the overlay just below your webcam. This way, when you glance at the answer, it looks like you are making direct eye contact. Most users report that interviewers cannot tell the difference.

Under the Hood

Tauri + Rust Backend

The entire backend is written in Rust, compiled to native binaries for Windows, macOS, and Linux. Memory usage is under 50MB. No Electron bloat, no Node.js runtime.

Deepgram Nova-2 STT

Deepgram's Nova-2 speech-to-text model provides the fastest and most accurate real-time transcription. Streaming WebSocket connection ensures sub-300ms latency from speech to text.

Multi-Model AI Layer

Six AI models available via a unified API layer. The app handles model selection, token management, and streaming responses. Answers appear word-by-word for the fastest perceived speed.

OS-Level Stealth

Window capture exclusion is handled at the OS level, not through CSS or opacity hacks. The overlay is genuinely invisible to screen-sharing APIs used by Zoom, Teams, and Meet.

How Stealth Mode Works