Friday evening. You want Monday's class to open with a short listening check on asking for directions, because that was this week's topic. Normally that means finding a clip that is almost right, recording yourself twice, trimming the audio, writing out a transcript, and then writing the questions. Two hours later you have one quiz. Most teachers quietly skip the listening part instead.
Quizzz does the recording for you. You describe the scene, pick the level, and the AI writes a short dialogue, voices it with two distinct speakers, and hands you a multiple-choice question that grades itself. You review it in the normal quiz builder before anyone hears it.
👉 Open Quizzz · Start free · Listening guide
Listening is a question type, not a separate tool
There is no separate listening editor to learn. Listening questions are three fixed audio formats you dial into the regular AI generation flow, in one of two ways:
- A dedicated listening quiz. Choose New Quiz and the Listening quiz: AI questions with audio tile. Pick the target language and level, set how many questions of each format you want, and describe the scenario in plain text.
- Listening added to any AI quiz. When you generate a quiz from a topic or from your own material, the same three-format picker appears at the bottom of the question mix. Whatever counts you enter are generated as extra questions on top of your reading and grammar items. Nothing in your existing mix is converted to audio.
The language picker only shows languages that have text-to-speech voices available, roughly forty of them. If a language is not listed there, it does not have listening support yet. The picker also does not appear on the Manual or Template tracks, because those never call the AI.
The three formats
The formats follow patterns that will be familiar to anyone who has sat a JLPT listening section, but they work for any language.
- Quick response. The student hears one short utterance and picks the most natural reply. Good for greetings, requests, and set phrases.
- Task comprehension. The student hears a short exchange between speakers A and B and decides what action should happen next. Good for instructions, plans, and sequencing.
- Key point. The student hears a short dialogue or explanation and extracts one fact: a time, a place, a reason, an item. Good for announcements and schedules.
You choose the mix. A beginner quiz might be six quick-response items and two key-point items. An intermediate quiz might lean on task comprehension.
A worked example: directions at the station
Here is the exact setup for a JLPT N4 practice set in Japanese.
- New Quiz, then the listening tile.
- Target language: Japanese. Level: N4.
- Formats: 3 quick response, 3 task comprehension, 2 key point.
- Scenario text: "Station announcements and asking a staff member for directions to the right platform. Keep vocabulary at N4."
Quizzz drafts a dialogue for each of the eight questions, synthesizes the audio, and opens the result in the builder. One of the key-point items might be a short announcement about a platform change, with the question asking which platform the train now leaves from. The four options are platform numbers. The student has to listen; there is nothing to read.
Every item arrives as a draft. An AI draft is never published on its own. You read the dialogue, listen once, fix anything you do not like, and only then share the quiz.
Editing the dialogue and the voices
Open any generated listening question in the content step and you can edit:
- the on-screen prompt the student reads before pressing play;
- each line of the dialogue: who speaks (A, B, or a narrator), what they say, and the pause before the line;
- the voice gender for each speaker, where the language offers more than one voice.
Audio does not re-synthesize while you type. When you are happy with the text, press Generate audio, or Update audio after an edit. That keeps you in control of when synthesis runs. The one limit to know: you cannot start a listening question from a blank line. The Add Question picker does not offer it, so hand-editing always starts from something the AI produced.
What the student experiences
Students open the quiz from a Magic Link, with no account and no app. For a listening item they see the format label, an audio player, and the answer options. The script is never shown and is never sent to the browser as text, so nobody can read the transcript instead of listening.
They can play, pause, seek, and replay as many times as they like. There is a Slow toggle that slows the narration without distorting the pitch, which lower-level students use a lot. Listening works in the Classic and Interactive answer modes. If you assign Guided mode to a quiz with a listening item, Quizzz falls back to Classic for that quiz, since Guided does not support audio-only items yet.
Grading is automatic, exactly like any other multiple-choice question, so results land in your dashboard with the rest of the quiz. You can also set the student interface language per share link, which matters when the quiz language and the students' first language differ.
Try one yourself: this picture-description listening demo is a real question from our own account.
Recently added
Three changes landed in the October release:
- Explanations and answer options can be shown in the learner's language instead of the target language, so a beginner can be tested on Japanese audio while reading English or Chinese options.
- You can upload your own audio for a listening question instead of using the synthesized voices, for example a recording of a real announcement or your own voice.
- A listening-only quiz can be generated from a PDF, grounded on every page of the document, so a textbook chapter can become an all-audio quiz.
Tips for good listening items
- Write a themed scenario, not a vocabulary list. "Ordering at a ramen shop and asking about allergies" produces tighter dialogues than "food words".
- Keep dialogues short. Two to four lines is enough for one fact. If a generated dialogue runs long, cut lines in the editor and update the audio.
- Match the format to the skill. Quick response for set phrases, task comprehension for instructions, key point for numbers, times, and places.
- Point lower levels to Slow. Tell students the toggle exists before the first quiz.
- Mind the cap. Every plan allows up to 20 listening questions per quiz, and the Free plan is capped lower. Short, frequent listening checks work better than one long test anyway.
- Listen once before you publish. Thirty seconds of review beats a confused class.
If you teach a language and have been skipping listening because producing the audio was the hard part, that part is now a text box. Create a free account, generate one eight-question set for next week's topic, and see how it sounds. Details are on the listening quiz maker page, with a Japanese-specific walkthrough on the JLPT listening test maker page and a step-by-step tutorial in how to create a listening test online.