Robot speech and languages in noisy event venues

A hired robot can speak clearly at almost any event; understanding visitors in a loud hall is much harder. Plan for robot-led, scripted interaction in a language the robot is confirmed to support, with a fallback – a touchscreen, a human host or an operator's voice – for when the room gets noisy.

Speaking is easy, listening is hard

Speech output is the reliable half. The robot reads a prepared script through text-to-speech, and you can test in advance exactly how it will sound.

Recognition is the fragile half. On Pepper and NAO, the NAOqi speech-recognition module recognises words or phrases from a predefined list and reports a confidence score for each match; it is based on speech technology from Nuance. In other words, the robot listens for expected phrases rather than understanding anything a visitor might say.

Pepper's user guide describes how this works in practice: Pepper detects people who come within about 1.5 metres, its eye LEDs show when it is listening, and visitors should say one short sentence in one go, because pauses are taken as the end of a sentence. If Pepper did not understand, it starts listening again and the visitor should repeat more loudly.

Some features need a network. Pepper's language documentation lists a separate remote recognition engine with a shorter language list, so ask which speech functions run on the robot and which rely on internet access – and test them on the venue's real Wi-Fi.

The Unitree G1 has a four-microphone array and a 5 W speaker, but Unitree's product page does not describe built-in voice languages. What a G1 says and understands therefore depends on the supplier's own software.

Which languages?

Check per robot, per software version and per unit.

Pepper. The NAOqi 2.5 documentation lists ten languages for text-to-speech and recognition: Japanese, English, French, Italian, German, Spanish, two Chinese variants, Dutch and Arabic.
NAO. The NAO V6 (NAOqi 2.8) documentation lists 21, including Czech, Polish, Russian, Swedish and Turkish.
Both pages warn that the lists are indicative and that support differs between languages.
Pepper usually has several languages installed and falls back to English if it cannot find a word or sentence in the current language, so a partly translated app can switch language mid-conversation.
Each language is separate content. In Choregraphe, every supported language has its own translation file to fill in and validate.

For a Czech or multilingual audience, ask the supplier for a short recording of the actual app, in each language, on the unit you will get. See the Pepper and NAO profiles for more on each robot.

Noise: what to expect

Event halls are loud, and venue rules limit how loud you may be in return.

A simple test. The UK Health and Safety Executive lists having to raise your voice to hold a normal conversation with someone about two metres away as one sign that noise at work may need action. If your stand staff have to do that, assume voice recognition at the robot will struggle.
Venue limits. Messe Frankfurt's presentation rules cap volume at 70 dB(A) at the edge of the stand and require an audience area at least 2 metres wide inside the stand in front of presentations, especially those using microphones. You cannot simply turn the robot up.
Small built-in speakers. Pepper's specification lists four omnidirectional microphones on the head and two loudspeakers in the ears with 7 W RMS output; the G1's speaker is rated at 5 W. For a stage or keynote, ask whether the robot's audio can be fed into the venue's PA (see conference use).

Placement helps: keep the robot away from PA speakers and busy aisles, give it the quieter corner of a trade-show stand, and make sure a visitor can stand close and face it.

Designing conversation for a noisy room

Let the robot lead. It asks a question with two or three possible answers and listens only for those words.
Write for the ear. SoftBank Robotics' design guide for Pepper recommends short sentences, everyday language and pauses, and suggests putting lists of more than three items on the tablet.
Always offer touch. The same guide recommends a voice trigger and a matching tablet button for every user input, so visitors can use either.
Tell visitors how to talk to it. A small sign ("Step closer and say one of these words") saves frustration.
Keep a human in the loop. A host can repeat questions for the robot or answer for it. A teleoperated voice, where an operator speaks live through the robot, handles open questions – be honest with visitors about it.
Mind privacy. Microphones can pick up bystanders; see cameras, microphones and privacy.

Checklist

Languages confirmed in writing for the exact unit and software version
Demo recording in each language, including brand and speaker names
List of speech functions that need internet, tested on the venue Wi-Fi
Expected noise level and the venue's volume limits checked
Robot placed away from PA speakers and aisles, with room for visitors to come close
Touchscreen or host fallback for every interaction
Audio feed to the PA arranged for stage moments
Script reviewed for short sentences and clear options
Languages and interaction style stated in your request

FAQ

Can a robot understand visitors at a busy trade show?

Sometimes, for short, expected phrases spoken close to the robot; for free conversation in a loud hall, not reliably. Pepper's built-in recognition works from a predefined phrase list, so plan robot-led dialogue with a touchscreen or host fallback.

Can a hired robot speak Czech?

It depends on the robot and its software. The NAO V6 documentation lists Czech for speech and recognition, while Pepper's NAOqi 2.5 list for those functions does not include it, and for a G1 it depends on the supplier's software. Ask for a recording from the actual unit.

How loud can a robot be on a trade-show stand?

Only as loud as the venue allows. Messe Frankfurt, for example, limits presentations to 70 dB(A) at the stand edge, so placement and clear scripts matter more than volume.

Should the robot's voice go through the PA system?

For stage moments and keynotes, usually yes, because built-in speakers are small. Ask the supplier whether the robot's audio output can be connected to the venue's sound system.

Sources

How this guide was made: prepared with AI assistance; every factual statement was checked against the sources listed above on 12 October 2026. Spotted an error? Write to info@humanoid4rent.com.