interactive ai avatars for events
An AI avatar as a conversation partner.
Not a chat window.
Aura51 puts an AI avatar on your stage as a discussion partner: on a panel, at a customer event, in a town hall or a lecture hall. It brings the voice that is otherwise missing there: disagreement, a follow-up question, an explanation or the view from outside. Not a pre-produced video: it listens, takes a position and speaks when the moderator calls it by name. It discusses in standard German; English is not available yet.
01Characters
Four stances. One system.
All four are AI avatars, not people. The same technology in all four; what sets them apart is stance, face and voice. You choose who takes the seat for each event.
02Use cases
The role your stage is missing.
Five formats to start with. The character stays the same; what changes is the briefing: you set the topic, the participants and the position.
Panel
The second view
Two or three guests and an AI avatar that takes the position missing from the panel.
Recommended: Albert, Cassandra
After a talk
Thesis and rebuttal
In the round that follows, it takes the opposing position when called on.
Recommended: Albert, Gustav
Town hall
The objection without a name
It raises objections submitted anonymously in advance, and leadership answers.
Recommended: Albert, Cassandra, Gustav
Customer event
Where does that come from?
It asks the questions your customers would otherwise only ask during the break, and asks for evidence.
Recommended: Cassandra, Gustav
University
Explain and follow up
It explains, and students follow up without saying its name.
Recommended: Ophelia, Albert, Cassandra
For a first engagement we recommend the panel.
03How it works
You are in charge.
Your moderator calls on the AI avatar, reads along and halts it with one tap. It says nothing on its own.
Example · the moderator’s tablet
Call
“Albert, how do you see it?” It speaks when called by name or at the press of a button, then briefly to follow-up questions, never on its own.
Read along
The moderator reads along on the tablet, usually a good second before the room hears the content. Nothing needs to be approved.
Stop
One tap on STOP and the AI avatar falls silent in under half a second.
What does not work.
If your first rule is “make nothing up”, you start with yourself.
- About two and a half seconds to the first word, about three to the substance, because it waits until the question is over: no rapid-fire exchange.
- It understands and speaks standard German, neither Swiss German nor, for now, English.
- Only with an internet connection.
- Instructed not to invent numbers: if the avatars do not know something, they say so.
04Measurements
Measured, with people in the room.
We measured live on the laptop, with people in the room and in natural speech. Every number carries its evidence level.
What we measured is the technology, and it is the same in all four: speech recognition, the call, the tablet and the stop. We check the character you book, with its voice and face, in the dress rehearsal on your system before the event.
- 2–3 sto the first sound, median about 2.5 smeasured · live
- under 0.5 sto silence after STOPmeasured · live
Not yet measured live: Two rules added after glitches · Summarizing long discussions
All measurements and where they come fromWhat is not yet proven
05Frequently asked
What organizers want to know first.
Does it just talk whenever it feels like it?
No. Only when called by name or by button. After that it takes follow-up questions even without its name, as long as the conversation keeps going without a longer pause, one minute at most. Then it needs its name again.
What if it says something wrong?
The moderator reads along on the tablet, usually a good second before the room hears the content, and stops it with one tap in under half a second. Nothing needs to be approved. If need be, a German “Danke” with its name stops it too; it falls silent after about a second.
Does it know who is speaking?
Yes. The moderator speaks into a dedicated microphone, and the guests’ names are in the briefing. We match the guests’ voices at the event once each guest has spoken a few sentences; anyone who gave a voice sample beforehand is recognized in the process.
What languages do the characters speak?
All four understand and speak standard German, neither Swiss German nor, for now, English.
Who runs the technology at the event?
We do. One of us sits next to your sound engineer, brings the laptop and the tablet, and can also stop the AI avatar from the laptop. Your technicians provide the moderator microphone as its own channel, the other stage microphones together on a second one, the screen and the return path into the desk. It also needs internet; we settle which line with your technicians.
Where does what is said on stage go?
The transcript stays on our laptop. At the event, four cloud services take part: speech recognition, two language models and the avatar service. We record a voice sample only with consent, and the voice profile made from it stays on the laptop. The services and where they are based, the data processing agreement and the deletion period are set out in writing with you before the event.
What does an engagement cost?
We do not publish a price list. What an engagement includes is listed under Request below; we share the terms after a short conversation.
06Request
Who do we bring onto your stage?
Tell us about your event.
An engagement includes the character with its briefing, the laptop, the tablet and our own network, one of us on site, coordination with your sound engineer and the dress rehearsal. We share the terms after a short conversation.
- 01
Request
You write to us: event, date, preferred character.
- 02
Coordination
We settle the briefing, the character and the technology with you.
- 03
Lead time
About two weeks to coordinate with your event technicians.