These two items defeat students who speak beautifully about nothing. The algorithm scores content by looking for the words that belong to the material in front of you — so naming things is the job.
Describe image. Twenty-five seconds to look, then forty seconds to speak. Use a fixed frame so you never have to invent structure under pressure:
- What it is: "The bar chart compares…"
- The biggest feature: "The most striking figure is…"
- One or two specifics, WITH numbers and labels: "Sales in the third quarter reached 40 million,
- roughly double the first quarter."
- A closing line: "Overall, the trend is upward across the period."
Say the axis labels, the units, the categories, the highest and lowest values. Those are the content words the software is listening for.
Retell lecture. You hear a ninety-second lecture, then speak for forty seconds. Take notes in two columns — the topic on the left, examples and numbers on the right — and begin with a frame sentence: "The lecturer discussed …, and made three main points." Then deliver the points using the speaker's own key terms. Reusing the lecturer's terminology is not cheating here; it is the content score.
For both: keep talking until the timer stops or you have genuinely finished. Silence scores zero for the seconds it occupies, and an imperfect sentence scores more than a perfect pause.