Speaking practice: record & get AI analysis

The speaking section has 26 tasks in five groups: personal questions, role-play scenarios, topics, photos to describe, and ⭐ advanced topics. Each task page gives you the question, key vocabulary, useful phrases, a self-check list, and a hideable model answer. Tap "Listen to the examiner" to hear the task read aloud, as in the exam.

Recording your answer

• Record straight in the browser (up to 3 minutes) — your browser will ask for microphone permission the first time.
• Play your recording back before submitting, and re-record as many times as you like.
• Mic not working? There's a file-upload fallback — record on your phone and upload the file (up to 10 MB).

What the AI analysis contains

• A transcript of what you actually said.
• A score, plus task completion — which points you covered and which you missed.
• Grammar errors with the rule explained, vocabulary level with suggestions, and a fluency comment — explanations in Portuguese and your interface language.
• A model answer to compare against.

The pair task: a second candidate

On the real exam you speak in pairs with another candidate. The section has 10 "Interação entre Candidatos" tasks — planning a weekend, choosing a restaurant — where the app plays Candidato B. The examiner reads the task, B opens, and you answer out loud; B reacts to what you say and helps you reach an agreement, in up to 8 turns. If the AI partner isn't available, a recorded partner carries on.

When the conversation ends you get feedback on your part straight away, like your other speaking answers, and it's kept in the task's history. You can still practise a task with a friend or tutor and mark "I practiced this ✓".