An AI examiner that follows up, like a real one
You speak, it listens and answers back in real time. When an answer is thin it asks why. When you ask, it repeats the question. In Part 2 it keeps the clock.
What a follow-up sounds like
Is it easier to learn something like cooking from a person or from an online video?
I think from a person. Because they can show you.
One line, no reason given, no example.
Why do you think a person is better at showing you than a video?
Instead of moving on, the examiner asks why, using your own words.
Well, a person can watch what you do and stop you when you make a mistake. For example, when my aunt taught me to cook, she noticed I was cutting the onions too thick. A video can't do that.
The second answer has a reason and an example. This is what Part 3 is listening for.
An examiner, not a chatbot
It does what a real examiner does, and nothing else.
没听清，可以再问一遍
和真实考官一样。提出请求，就能原封不动再听一遍。
你·Sorry, could you say that again?
考官·Sure. Which local place do you miss when you are away?
回答太短，就会继续追问
Part 3 只答一句，会被追问“为什么”；充分展开，就进入下一题。
你·I think from a person. Because they can show you.
考官·Why do you think a person is better at showing you than a video?
Part 2，严格计时
最多讲两分钟。如果沉默 6 秒，考官就会结束这一部分，继续往下考。
英音或美音，你来选
选择你想在考场上听到的口音。两种声音使用同一套题库。
When it speaks, and when it doesn’t
| Talks in real time | A spoken conversation with an AI examiner. You answer out loud; there is nothing to type. |
|---|---|
| Follows up | When a Part 3 answer is thin, it asks why or asks for an example. A developed answer moves on to the next question. |
| Part 2 | 1 minute to prepare with notes, up to 2 minutes to talk. 6 seconds of silence ends your turn, as in the real test. Then 1 short question. |
| Part 3 | Up to 2 follow-ups on one question, linked to your Part 2 topic. |
| Repeats | Ask and it repeats the question, word for word. |
| Stays neutral | It doesn’t praise, hint or correct you during the test. Feedback comes in the report afterwards. |
| Voice | British or American examiner voice. |