Host a multilingual meeting

Translated meetings fail in predictable ways. The microphone picks up the wrong voice. The QR code is too small to scan from the back row. A presenter steps away from the mic mid-sentence. None of these are technology problems — they are room problems. This guide is the operator’s checklist for avoiding them.

Choose the right microphone

Realtime speech recognition is forgiving of accent, cadence, and vocabulary. It is unforgiving of poor audio — see audio requirements for the full technical baseline. Three rules:

1. One speaker, one microphone. Lavalier or headset microphones are ideal — they travel with the speaker and maintain consistent gain. For a detailed comparison, see Choose the right microphone. If a table or boundary microphone is the only option, define a single “speaker seat” and brief participants on it.
2. Hardwire whenever possible. Bluetooth microphones introduce variable codec quality and occasional dropouts. USB or 3.5 mm wired connections are reliable.
3. Test before guests arrive. Open a throwaway session 10 minutes before start. Walk the room. Read a paragraph from the agenda and check the live transcript on the presenter view.

Pick the join surface

The QR code is the primary join surface. Three placements work, in order of preference:

• Projected on the main screen before the session begins. Visible to everyone, low-friction.
• Printed and placed at each seat. Useful for fixed-seating rooms and high-stakes meetings where the projector is reserved for the agenda.
• Embedded in the calendar invite. The least reliable — attendees often join with the calendar event closed.

Always have the session code displayed alongside the QR. There will always be one attendee whose phone camera will not focus.

Brief the room

Thirty seconds at the start of the meeting buys an hour of smooth running. If any speakers are presenting in a second language, share tips for non-native speakers with them before the session.

“We are using realtime translation today. Scan the QR or enter the code on your phone, choose your language, and hold the phone close. Subtitles appear on screen; audio plays through your phone — use earbuds if you are close to the speaker.”

That last point matters. A phone playing translated audio next to the original speaker creates an echo that confuses both the room and the speech recognizer.

Watch the transcript

The presenter view shows a live transcript of recognized speech. Keep it in your peripheral vision. If the transcript drifts — missed words, garbled phrases — it is usually one of three things:

• The speaker has stepped away from the microphone.
• Two people are talking at once.
• The room has gained ambient noise (HVAC kicking on, a door propped open onto a corridor).

A short pause and a return to the microphone resolves nearly all of these. The recognizer recovers within a sentence.

After the session

End the session deliberately by pressing End session, not by closing the tab. This finalizes the transcript and triggers the export build. Download the multi-language transcript immediately — it remains available, but the moments right after a meeting are when context is freshest if you need to attribute a quote or follow up on a question. For a full cleanup workflow, see Curate transcripts after the event.