A listening lesson with AI-generated audio, built around your learner's job
Coursebook recordings rarely talk about your learner's job. With AI you can prepare a short audio dialogue in their sector and at their level. Here is the method, in four steps.
In brief
A tailored listening activity is built in three stages: a short script, one to two minutes long, written with AI from a real situation in the learner's job and at the target CEFR level; voices generated with a text-to-speech tool, varying speakers and accents; and three listenings in the lesson: gist, specific information, then language. The trainer rereads the script and listens to the whole audio before using it.
1. Start from a real situation
The starting point is the needs analysis: who does the learner speak to, through which channel, about what? LangueXpert certified trainers have prepared, for a director of sport, a press conference facing journalists in three different English accents; for a cloud engineer, audio dialogues in his sector; for a doctor, audio materials built on the situations she described in the very first session. Their accounts are on the certification page.
2. Write the script at the right level
Ask the AI for a short dialogue between two or three speakers, using the vocabulary of the job, pitched at the target CEFR level. Ask for the comprehension questions and the answer key at the same time. Then reread it: an approximate technical term, or a sentence too long for the level, is fixed before the voices are generated, not after. Describe the job and the situation, without naming any person or company.
3. Generate the voices
Paste the script into a text-to-speech tool, give each speaker a voice and choose the accents the learner actually meets at work. Listen to the whole audio before the lesson.
4. Three listenings
- First listening: gist — who is speaking, about what, to what end.
- Second listening: specific information — dates, figures, decisions.
- Third listening, with the transcript: language — useful expressions, phrases to reuse.
The lesson ends with production: the learner replays the situation with their own information.
Example
A maintenance technician, from A2 towards B1: a customer calls to report a fault. A ninety-second script, two voices, one of them a customer with an American accent. Questions: what is the problem, since when, what solution is offered? Language to keep: describing a symptom, asking someone to repeat, offering an appointment.
A starter prompt
You are an English trainer for adults. My learner works as [job], current level [A2], target [B1]. Write a dialogue of about 200 words between [two speakers] in this work situation: [situation]. Use the vocabulary of the job and sentences suited to the target level, and put each speaker's name at the start of each line. Then add 3 gist questions, 5 detail questions, 6 expressions to remember, and a separate answer key.
The full version — the resource library, the lesson plan, the audio materials and how to use them in class — is built in the preparatory course of the certification, on one of your own learners. See also: the speaking assessment rubric and the placement test.
Frequently asked questions
Can I teach listening with a synthetic voice?
Yes, for practice targeted at the learner's job: you choose the situation, the level, the pace and the accent. These materials complement authentic recordings (meetings, calls, videos from the sector); they do not replace them.
How long should a listening audio be?
One to two minutes is enough: that leaves time for three listenings and a full follow-up in the lesson. Several short audios, one situation each, work better than one long recording.
Which tools do I need?
An AI assistant to write the script and the questions, then a text-to-speech tool for the voices. In the LangueXpert preparatory course, the tools worked on are Claude, ElevenLabs and Google AI Studio.
Can the AI get the script wrong?
Yes: an approximate technical term, a level that is too high, an unnatural turn of phrase. That is why the trainer rereads the script before generating the voices and listens to the whole audio before the lesson.