A phone at the lectern
A smartphone close to the speaker records the voice. Set up in minutes, right for receptions and smaller rooms.
Live translation, tailored to your event.
Made one event at a time. Before the day we study your speakers, their institutions and their expert vocabulary, and fit every step to your topic, your room and the way your speakers talk. On the day, your guests read each speech live, in their own language.
- Prepared for your speakers and expert vocabulary
- Tuned to accents and non-native English
- In your brand, on phones or big screens
- No audio recorded
Prepared for this event
Ama MensahKivu CapitalKigaliFrontier marketsNon-native speakers
Your event, your brand
Your guests see only your brand.
The reading page and the QR cards carry your identity. For every event we set everything up from scratch.
- Your colours, typeface and logo
- QR cards in the design of your invitation
- On your own web address, if you wish
G
Your GroupAnnual Conference 2026
Ladies and gentlemen, welcome to our annual conference.
Our first guest is Ama Mensah of Kivu Capital in Kigali.
Read live
G
Read the speeches live
Scan the code with your camera
Benchmark
Facundia made the fewest errors.
South Africa's finance minister in parliament, on how contracts were split up to avoid public tenders. 27 seconds, translated live into German by four tools. Each translation appears at the moment it arrived in the test.
The clip has ended. The translations are still arriving.
Said
When the text arrived
Wrong meaningImpreciseEnd of sentence
Live into German
Measured live on 6 October 2026, one run per tool, on 27 seconds of Enoch Godongwana's budget speech (CNBC Africa). Facundia had the occasion and a list of names, the others a plain instruction. Seconds from the end of a sentence to the translated text, median over five sentences. Errors as judged blind by two AI reviewers from different providers.
How it works
Assembled anew for every event.
Off-the-shelf models don't know your speakers, your expert vocabulary or your room. So we prepare each step for your event, and for each one we pick the best of several tested models.
Your speakers' accents
Your expert vocabulary
Your topic and tone
- 01ListeningSpeech to text, with the model that understands your speakers best
- +SpeakersOptional. When several people speak, we recognise each voice, so even a fast, open discussion reads cleanly
- 02CorrectionNames, expert vocabulary, slips of the tongue
- 03TranslationIn the context of the whole speech
- ■Your guestsPhone or screen
Test benchBefore every event, speeches like yours, with the same topic and the same accents, run through several models at each station. The best at each station goes live.
ListeningModel AModel BModel C
CorrectionModel AModel BModel C
TranslationModel AModel BModel C
At your event
BAC
The best of each stationSound
How the voice reaches us.
We're happy to meet your sound technician beforehand and find the best setup together.
Straight from your sound system
A dedicated output from your sound system delivers the voice clean, without room echo or applause. The best choice for halls and panels.
Use cases
Strongest where the topic is narrow and the stage is live.
Receptions
Welcome speeches and toasts, with no headsets in the room. Your guests read quietly on their phones.
Panels and podiums
Several voices, a moderator and questions from the floor. We recognise who is speaking and label each contribution.
Specialist conferences
Medicine, finance, law, research. The narrower the topic, the more the preparation counts.
Languages
All major languages, in both directions.
English to German, German to English and between all major languages. Where German is involved we are especially precise: long sentences with the verb at the end, numbers in the right format, titles and forms of address as the room expects them.
Privacy
Discreet, GDPR-compliant, on audited infrastructure.
Servers in Germany
Pages and texts are hosted by Hetzner in Falkenstein, in a data centre certified to ISO/IEC 27001.
Audited AI providers
Our providers for speech recognition and translation are audited to SOC 2 Type II and do not train models on your content.
No audio stored
We store no audio. The sound is recognised, translated and discarded, encrypted via TLS throughout.
Discreet
Under a confidentiality agreement on request. If you wish, we delete the text after the event.
FAQ
Frequently asked questions
Do my guests need an app?
No. They scan a QR code or open a link and read along in their browser.
Why is your service better than the others?
In the end, we all use the same AI models. Our approach is artisanal. For each event we try many combinations of models and keep the one that works best for your speakers and your topic. Self-service tools give everyone the same off-the-shelf setup, and to save costs they often leave out the best model.
What does it cost?
From €1,000 per event, fully tailored to your occasion. For every event we set up the software anew from our own building blocks. That takes manual fine-tuning and a review by experts.
How quickly does the translation appear?
After a few seconds. It depends on the accent and on which language is translated into which. We translate sentence by sentence, because every language orders its words differently. Depending on the sentence, the translation waits until the speaker has finished it.
What if a speaker has a strong accent or limited English?
We prepare for it. Before the event we test speech recognition models on recordings with the same accent and choose the one that understands your speakers best. The translation renders what the speaker meant and smooths out broken grammar. If the sound is poor, it waits a little longer for whole sentences, which costs a moment and keeps the sentence structure right.
What if speech recognition mishears expert vocabulary?
It often happens with rare terms. Before the event we collect the expert vocabulary of your field in a glossary. The translation knows it and everything said so far, and puts in the right term.
What do I need on site?
A sound input, for example a smartphone at the lectern, or a connection to your sound system. An internet connection over Wi-Fi or mobile data. And for your guests, QR codes on the tables or a screen in the room.
How do we connect the sound?
Through a microphone near the speaker, which can be a phone, a laptop or any mic, or directly from your mixing desk. We're happy to speak with your technician beforehand.
Which languages are possible?
All major languages in both directions, for example English to German or German to English. Ask us about your combination.
Is the sound recorded?
No. We store no audio. The sound is recognised, translated and discarded.
Your next event?
Send us the date and topic. We'll get back to you, prepare your glossary and set up your access.