NewPowered by gpt-live-1
A real examiner that listens while it speaks
Your interview runs on OpenAI's gpt-live-1, released in September 2026 — a full-duplex voice model, so the examiner hears you even mid-sentence. Real 3-part format, adaptive follow-ups, and a practice band with feedback at the end.
The real 3-part format
Part 1 interview, a timed Part 2 cue card and a Part 3 discussion — with follow-up questions that adapt to what you just said.
Full-duplex turn-taking
gpt-live-1 listens while it speaks, so you can pause to think, interject, or ask for a question to be repeated without talking over a robot.
A properly timed cue card
One minute to prepare, then your long turn — say “I’m ready” whenever you want to start early, exactly like the test-day script.
Band estimate and feedback
When the interview closes you get a practice band with criterion-by-criterion feedback and what to practise next.
How a session goes
- 1Examiner greets you
- 2Say your name for an audio check
- 3Part 1 questions
- 4Part 2 cue card — 1 min prep
- 5Part 3 discussion
- 6Automatic end and your band