A multilingual Q&A works when three things hold: every question is asked into a microphone, that microphone feeds the translation, and the speaker can read or hear the question in their own language before answering. Most Q&A sessions that go wrong in several languages do not fail on translation. They fail on sound: a question called out from the third row, a microphone that only feeds the room speakers, a panelist who nods politely without having understood. Here is a practical setup, step by step.
Why the Q&A is the hardest part to translate
A keynote is predictable. One person speaks into one microphone, often from slides prepared in advance. The Q&A is the opposite. Several people speak in turn, from anywhere in the room. They switch language without warning. They start talking before the microphone reaches them. And the answer has to come back to them in a language they understand, or the exchange falls flat.
With human interpretation, the interpreters in the booth hear the question through the room audio and switch direction as needed. With live AI translation, the principle is the same, and so is the constraint: a system can only translate the sound it receives. A question that never reaches a microphone does not exist for the translation, whatever tool you use.
Step one: decide where the sound comes from
Before the event, ask a simple question: what feeds the translation? With Glot, the sound comes from the room's mixing console, from a wireless lavalier microphone, or from a USB table microphone. Each option shapes how you run the Q&A.
If you take the sound from the console, everything the technician mixes for the room can be passed on: the speaker's lavalier, the table microphones, and the roaming microphones handed to the audience. This is the most robust setup for a Q&A, because audience questions follow the same path as the speaker's voice. Check with the technician that the roaming microphones are routed to the output you use, and not only to the room speakers.
If you take the sound from a single lavalier worn by the speaker, it will capture the speaker very well and the audience very badly. That is fine for a talk, not for a Q&A. In that case, plan for the moderator, or the speaker, to repeat each question into their own microphone before answering.
If you use a USB table microphone, which suits small rooms and workshops, the same logic applies. It hears the people sitting close to the table. A question from the back of the room will be faint or lost.
Step two: put a microphone in every questioner's hand
The rule that makes a multilingual Q&A work is simple: nobody asks a question without a microphone. It sounds obvious. In practice, it is the rule most often broken, because people are impatient and the runner is still on the other side of the room.
A few habits help. Announce the rule at the start of the session. Have one or two runners with roaming microphones, and ask people to raise their hand and wait. Ask questioners to give their name and to speak at a normal pace. If someone starts talking anyway, the moderator can stop them kindly and ask them to start again once the microphone arrives.
Another approach works well: collect written questions. Attendees send them during the talk, through a form or a chat, in the language they prefer, and the moderator reads a selection aloud. The moderator's voice is clear and close to the microphone, and each question is spoken once, cleanly. You lose some spontaneity, but you gain reliability, and it suits large rooms where passing a microphone around takes time.
Step three: make sure the speaker understands the question
This is the part organisers forget. The audience follows the talk in their own language, but who translates the question for the speaker? If an attendee asks a question in Portuguese and the speaker only understands English, the speaker needs the translation too.
With a phone based setup, the answer is simple: the speaker scans the same QR code as the audience. Attendees follow in the browser of their phone, with no app to install and no account to create, and the speaker can do the same at the lectern or in their chair. They pick their language and read the live captions, or listen to the translated voice through an earpiece if you use the Voice plan. Live captions on the room screen are another option, as long as the speaker can see the screen from where they stand. We cover that setup in our guide to live captions on the big screen.
Leave a short pause after each question. Live translation takes a moment: with Glot, the translation arrives about two seconds after the words are spoken. A speaker who waits a breath before answering gives everyone time to catch the end of the question.
Step four: help everyone know who is speaking
In a Q&A, the voice changes every minute. For someone reading captions, it is easy to lose track of whether a sentence belongs to the speaker, the moderator or someone in the audience. Captions that show who is talking make the back and forth much easier to follow, on a phone or on the room screen. The moderator can help by naming people as they go: a question from the second row, over to you, Maria.
Step five: rehearse the Q&A, not just the talk
Technical checks usually cover the speaker's microphone and the slides. Add a few minutes for the Q&A. Walk to the back of the room with a roaming microphone, ask a question in another language, and check on a phone that it appears in the captions. Then check that the speaker can read it from where they will stand. That test catches most routing problems before the audience arrives.
What to tell the audience at the start
A short announcement saves a lot of confusion. Tell people they can follow the session in their own language by scanning the QR code on the screen. Tell them they are welcome to ask their question in the language they are most comfortable with, as long as they use the microphone. And tell them that people who prefer the original can listen to it directly, since the translated voice is an option, not an obligation.
After the session
The Q&A is often where the most useful content of a session comes out: the objection nobody had anticipated, the practical example, the follow up the speaker had not planned. Glot produces a written summary of the session at the end, a handy record for attendees who missed part of the exchange. We explain why that matters in our article on AI session summaries.
A multilingual Q&A does not need a booth or a stack of headsets. It needs a clear sound path, a microphone discipline the whole room understands, and a way for the speaker to follow the question. Get those three right and the conversation flows in every language in the room.
Want every question in the room understood, in every language?
See how Glot works