It hears how you sound — and never keeps the recording.
You cannot hear yourself the way an interviewer does. Every spoken answer is measured while it is fresh: how fast you went, how many fillers slipped in, where you stalled, and how the voice itself came across. Then the audio is deleted.
What gets measured
- Answer length — and which way you missed: too short, too long, or well judged.
- Speaking pace — words per minute against the band that is easy to listen to.
- Fluency — fillers per minute, with long pauses counted too. Two "um"s a minute is human; a dozen is what interviewers notice.
- Confidence — heard in the voice, not guessed from the words.
- Clarity — how easily the words would be made out. An accent is not a defect and never lowers this; it measures articulation, not nativeness.
Delivery lifts a real answer — it cannot replace one
A fluent, confident, beautifully paced answer to the wrong question is still the wrong answer, and it is scored that way. Delivery adds up to a quarter on top of what you actually said, and nothing on top of nothing. That is deliberate: the alternative rewards polish, and polish is not what gets you hired.
What we could not measure is reported as "not measured" rather than as a zero — a written exercise has no pace, and telling you your clarity is 0/100 when we never heard you would be a lie.
The interviewer reacts to it like a person
If you sound tense, you may be told to take your time — the way a decent interviewer actually does. What you never get is a clinical read-out mid-interview: nobody in a real room says "your jitter proxy is elevated". The numbers are for your debrief afterwards.
The privacy part, plainly
Your answer is transcribed and analysed in the moment, and the audio file is deleted — we keep the transcript and the measurements, not your voice. There is no self-view and no video of you anywhere; the app never requests camera access at all. The video in an interview is the interviewer's, rendered by us.