Live translation for conferences, read on attendees’ phones
Waavoo uses AI to translate a conference live and displays the translation as text, on each attendee’s phone or on the room’s screen. People open a link or scan a QR code: no app, no account, no interpreter booth. Seven languages are available, one translation language at a time per session.
How does a live-translated conference work?
The speaker talks as usual, into the room’s microphone. Nothing changes on stage: everything happens between the tech desk and the audience’s phones.
1. Set up the session
In the console, the operator picks the speaker’s language and the translation language. Starting in the first quarter of 2027, they will also be able to upload context documents (a glossary, the program) to improve accuracy on names and field-specific terms.
2. Connect the sound
The console runs in a web browser on a laptop, at the tech desk or by the stage. It listens to the laptop’s audio input: ideally the output of the mixing desk, through a USB audio interface.
3. Share the link
The session’s QR code appears on the room’s screen. Attendees scan it and read the translation on their phones as the speaker talks. The link can also be sent out, for instance to people attending remotely.
What does the audience see?
A single web link. No download, no account. Every viewer reads the session’s translation in real time, with typography and contrast tuned for the back row.
- On their phone: five reading themes, including a high-contrast one, and text size adjustable from 24 to 150 px.
- On the room’s screen: an “ultra-clean” projection mode with no interface, just the translated word. The QR code can be brought back with a single key for latecomers.
- Remotely: the same link. Someone following the event online reads the same translation as the room.
The viewer is built to the relevant WCAG 2.1 AAA criteria: caption contrast is at least 8.3:1 in each of the five themes. The details are in our accessibility statement.
Which languages are available?
Seven languages recognized at the source, seven translated on screen: French, English, German, Spanish, Italian, Portuguese and Dutch.
A session shows one translation language at a time, set by the operator, who can switch it mid-session. To serve two languages at once (an audience that is part English-speaking, part Spanish-speaking, say), you need two simultaneous sessions. Tell us which languages you need and when, and we will confirm availability.
The translation language can also be the speaker’s own: French to French, for example, with French set as the speaker’s language. The session then shows real-time captions in that language, useful for deaf and hard-of-hearing attendees.
What do you need in the room?
Our recommended setup:
- A laptop running Chrome or Edge, for the operator’s console.
- The room’s sound: the mixing desk output, connected to the laptop through a USB audio interface set as the default audio input. Failing that, a microphone close to the speaker.
- A stable internet connection for that laptop, wired if possible.
- For the audience, nothing but their phone and a connection: the venue Wi-Fi or mobile data.
- A test run half an hour before you start, in the room, with the real microphone.
A remote speaker joining by video call must go through the room’s sound system: the console only captures the laptop’s audio input, not the sound the laptop plays.
When is an interpreter the better choice?
Waavoo lets you do without a booth when the audience mainly needs to follow the speaker. It is not the right fit everywhere:
- Two-way exchanges. Waavoo translates what goes into the room’s microphone, into one language. If audience members must speak in their own language and be understood by the speaker, you still need interpretation.
- Diplomatic, legal or medical settings, where every nuance carries weight: a professional interpreter answers for their translation.
- Attendees whose first language is a sign language: captions do not replace an ASL, LSQ or LSF interpreter.
We do not publish an accuracy figure yet. We will only publish a complete measurement, taken on the production pipeline and reviewed by a human evaluator. Until then, judge for yourself: ask us for a demo using an excerpt of your own content.
What happens to the audio and the text?
Speech-to-text and translation run on our own AI models: no third-party AI service processes the audio. Audio is never recorded, captions and translations are deleted at the end of the session, and nothing is used to train models.
Our main servers are in France. Backup compute servers in Montréal only take over speech-to-text and translation when the French ones are unavailable. The full details are in our privacy policy.
Frequently asked questions
Do attendees need to install an app?
No. They open a web link or scan the QR code shown in the room: no download, no account. A recent web browser is all they need, on a phone, a tablet or a laptop.
How many languages can be translated during a conference?
One translation language at a time per session, out of seven: French, English, German, Spanish, Italian, Portuguese and Dutch. The operator can switch languages mid-session. Serving two languages at once takes two simultaneous sessions, subject to availability. Just ask us.
How many attendees can follow the translation?
Audience size does not change the price: each attendee reads on their own phone. For a large audience, let us know ahead of time and we will prepare the session with you.
What internet connection is needed?
A stable connection for the operator’s laptop, which sends the sound to our servers, wired if possible. Attendees use the venue Wi-Fi or their mobile data; they only receive text.
Can the translation be shown on the room’s screen?
Yes. Projection mode shows the translation in large type with no interface at all, on the room’s screen, from a browser on the computer driving it. The session’s QR code can be displayed there too, so the audience can join from their phones.
Is the conference audio recorded?
No. Speech is turned into text in real time and discarded immediately; no audio file is ever created. Captions and translations are deleted at the end of the session.