Interview Copilot · live & screen-share safe
Real-time interview answers,
grounded in your prep.
Interview Copilot hears the question on-device, retrieves from your résumé and the company's Q&A database, and streams a speakable answer to an overlay only you can see. Behavioral, system design, or domain: the answer sounds like you and fits the role.
10 min/day free · no card · Windows & macOS
The pipeline
It hears, it grounds, it says. In about a second.
The whole loop runs the moment the interviewer stops talking. No clip uploads, no round-trip latency, nothing to catch up on.
Hears
A fully on-device Vosk recognizer transcribes the interviewer live. Audio never leaves your machine, so there is no upload lag and nothing to intercept.
Grounds
Your résumé and the company's Q&A database are retrieved and fed to the model, so the answer draws on your real experience and the role, not a generic script.
Says
A tight, first-person answer streams into the overlay, ready to speak. The prepared answer auto-scrolls into view as the question lands.
Inside the copilot
Everything that makes the live answer land.
Six systems working together, from the microphone to the words on your screen.
On-device Vosk STT
A streaming Vosk recognizer transcribes the interviewer locally in real time. No clip uploads, no cloud speech API, just audio that stays on your machine.
RAG grounding
Retrieval pulls from your résumé and the company Q&A set before the model answers, so every response cites your real experience and the company's domain.
Auto-scroll to your answer
When a question matches your prepared bank, the copilot scrolls straight to that model answer, so you glance down and it is already there.
Multi-model, in parallel
Answers run across Grok, DeepSeek, and Gemini at once, with a local model option for a fully offline path. You pick which models get called.
Screen-share invisibility
The overlay opts out of capture at the operating-system level, so Zoom, Teams, Meet, OBS, and PrintScreen never record it.
Global hotkeys
M pauses the copilot instantly; P toggles ghost mode to dim the overlay. Both are global, so you never touch the window mid-call.
Company-specific prep
Pick a company. Get a Q&A database built for it.
Behavioral, fundamentals, and coding questions with model answers in your voice. The copilot grounds every live answer in this set.
Questions
How the live copilot works.
How does it answer in real time?
An on-device Vosk recognizer transcribes the interviewer as they speak. The moment they stop, your résumé and the company Q&A database are retrieved and fed to the model, and a speakable answer streams into the overlay in about a second. No clip uploads and no round-trip latency.
Can the interviewer see the overlay on screen share?
No. On Windows the overlay uses an OS-level screen-capture exclusion, so it renders on your monitor but is composited out of the frame Zoom, Teams, Meet, Discord, OBS, and PrintScreen capture. It also stays out of the taskbar and Alt-Tab, and a live shield badge shows the current protection state.
What are the answers grounded in?
Every live answer is retrieved from your résumé plus a company-specific Q&A database of behavioral, fundamentals, and coding questions with model answers in your voice. You build that set in seconds with the Company Q&A Generator above.
Which AI models does it use?
It queries Grok, DeepSeek, and Gemini, and can also run against a local model for a fully offline path. You choose which models to use, and only the ones you pick are called.
What do the hotkeys do?
M pauses the copilot instantly so it stops listening and answering; P toggles ghost mode to dim the overlay. Both are global hotkeys, so you never touch the window during the call.
Walk in with the answers ready
Bring the copilot only you can see.
Start free in a minute. Build your company's Q&A set today.