Audio and device setup for in-person events

How to place devices and prepare rooms so in-person discussions come out clearly — with a checklist to run before the event.

One recording device per discussion group, placed in the middle, in a reasonably quiet room — that setup covers most events. Run one test with the exact device, browser and Wi-Fi you will use on the day; that single test prevents most audio problems.

Prefer the visual guides?

This article is the quick text version. For more depth:

Set up audio and devices — the video lesson for setup in the deliberAIde Academy (sign in to watch): every recording, in the room or with guest phones at the tables, online calls, during the session and when something goes wrong, and a one-minute test recording.
Audio & Tech Setup Guide (PDF) — an illustrated, printable step-by-step guide to audio and device setup for organisers.
Also in the Academy: Record a discussion live and Invite guests with links and QR codes.

You also find them in the app at any time: click the ? button, open Help, and look under Guides & tutorials.

Which setup fits your group

You can record directly in the platform on a laptop, tablet or phone, or let groups record with their own phones through guest links.

Group size	Recommended setup
1–3 people	Place a phone between the participants, or pass it to whoever is speaking.
4–15 around a table	Put one phone, tablet or laptop in the centre of the table. A 360° table or conference microphone improves the sound.
15+ in one room	Can everyone hear each other without microphones? Then one central device may work — test it first. If not, take a feed from the sound system, or use microphones: handheld microphones passed around, or conference microphones. Or split into table groups, each with its own device. Several tables in one room work fine.
50+ (large venue)	Let each breakout table record separately. For plenary parts, take a feed from the venue's sound system, or use handheld or conference microphones.

Where to put the device

Put the device in the middle of the group, at about the same distance from everyone, and within about 1.5 m of each speaker. If someone sits further away, pass the device to them or use a table microphone.
A phone lying flat on the table usually hears a group better than a laptop. A laptop's microphone picks up sound from the screen side, so face the screen toward the group.
Several tables in one room are fine: give each table its own device and its own link.
Wired and conference microphones, including a 360° table microphone, are fine. Don't plug headphones, earphones or a headset into the recording device, and don't connect Bluetooth headsets or speakers to it.
Is someone joining by video call? Play the call through the loudspeakers, or put a phone with the recording next to the laptop that runs the call. See Record online meetings.

Before the event

Make one person responsible for recording at each event. Recording while moderating is hard. Brief everyone who will record for a few minutes, or ask them to watch the setup video.
Do a full test recording with the real device, browser and internet connection, from where a group will actually sit. Read the transcript back: every voice from every seat should appear, not only the person nearest the device. For guest links, test on mobile data as well as Wi-Fi.
Live transcription needs a stable internet connection. No internet at the venue? Record with the phone's voice recorder app and upload the files afterwards.
Charge every device fully, bring chargers, and check that sockets are within reach of the tables.
On laptops, disable sleep mode — recording can stop when the device sleeps.
Open the recording page once before the event and allow microphone access when the browser asks.
For an important session, make a backup recording on a second phone with its voice recorder app. Delete it once the recording in deliberAIde is fine.

Phone and tablet settings

Switch off screen lock (auto-lock), or set the screen timeout to the maximum or "Never". A locked or dark screen records nothing.
Switch off low power mode (iPhone) or battery saver (Android). It can turn auto-lock back on and override the setting above.
Keep every recording device plugged in for the whole session.
Turn on Do Not Disturb, so calls and notifications cannot interrupt the recording.
On Apple devices, turn Bluetooth off so the microphone doesn't switch devices mid-recording. On MacBooks, set the microphone mode to "Standard" instead of Voice Isolation.

During the recording

Start recording straight after the microphone check, so the page or invite link doesn't sit idle in between.
On a phone or tablet, stay in the browser on the recording page the whole time. Switching to another app pauses the recording. On a laptop you can switch tabs or share your screen, but never close the recording tab.
For a short pause, such as a silent activity, use Pause. For a real break, such as a coffee break, click Stop, then Save and stop, and click Record again afterwards. On a guest link, the group taps Stop & Submit; after the break, click Reactivate same link and choose Continue on the same link.
If a warning appears above the recording, see Recording health warnings.

How accurate can a transcript be?

No in-person recording is perfect, so no transcript of one is flawless. In a room, every microphone also picks up other voices, echo and background noise, and people talk over each other. Unless every participant speaks into their own microphone (as in a video call, where each person has their own audio channel), perfect audio is practically impossible, and so is a spotless transcript.

Even the best speech recognition makes mistakes:

Clear speech: on clear recordings of one person speaking, today's best systems get about 1 to 3 words in 100 wrong in widely spoken languages such as English, German, Spanish or French.
Group conversations: in lively group conversations recorded with room microphones, error rates are far higher. In international research challenges on exactly this task, even the best systems still get about 1 in 5 words wrong or assign them to the wrong speaker, and about 1 in 3 in the hardest recordings, such as conversations around a dinner table.
The microphone: in one published comparison of the same meetings, a system made about twice as many errors with a single microphone on the table as with a microphone worn by each speaker (36 compared with 17 words in 100).
The language: speech recognition learns from recorded speech, and some languages have far more of it than others. Languages with a lot of training data are recognised much better than languages with little.

What deliberAIde does about it: after each contribution, deliberAIde recognises that passage a second time, more thoroughly, and corrects the errors it finds. Some errors still remain, so check quotes, names and numbers before you publish them.

Why it's still worth it: people speak about three to four times faster than the average person types, so a note-taker has to choose what to write down, and much is lost or paraphrased. A transcript, even with some errors, records the whole discussion, not just what someone managed to note down. Together with AI-assisted analysis, we believe this captures considerably more voices, ideas and perspectives than a note-taker alone could. Speech recognition keeps improving, and we update deliberAIde as it does.

Last updated: 2026-10-04

Wasn't helpful? Tell us what's missing