The four-stage pipeline
Listen
The app captures the interviewer's audio from your video call and transcribes it to text in real time.
Understand
It classifies the question — technical, behavioural or coding — and combines it with your resume and the job description.
Generate
A large language model produces a precise, personalised answer in the right style for that question type.
Display
The answer appears on your invisible overlay within roughly a second, ready for you to read or build on.
Real-time transcription
Speech recognition runs continuously while the interviewer talks, so there is no "record then process" delay. The transcription is tuned for interview audio — varied accents, fast speech and imperfect connections — so questions are captured accurately.
Context-aware answer generation
The assistant does not answer in a vacuum. It feeds your resume and the job description into the model alongside the transcribed question, so answers cite your actual projects and align with the role's real requirements instead of sounding generic.
The invisible overlay
The answer is shown on a desktop overlay rendered in a way that screen-capture and screen-sharing tools cannot record. It also stays hidden from Alt+Tab and the taskbar — the full mechanism is covered on the invisible overlay page.
The desktop app processes interview audio locally on your device — interview transcripts and answers are not stored on GirGit AI servers.