Instant Live Translation for Your Events

You receive a total of 30 minutes of real-time AI-powered translation to use in your browser—no credit card required. These minutes are cumulative, so you can test the system over several days: for example, try it a week before your event with colleagues, another day with your supervisors, and again on the event day for a final technical check. This way, you have the flexibility to evaluate our solution thoroughly in various real-world scenarios.

The trial uses the standard system, without event-specific preparation. See how we prepare the AI for your event.

The system stops working, but nothing is charged. You can contact us to upgrade and continue using the tool under one of our paid plans.

No. Nubart TRANSLATE runs entirely in the browser and works on all common smartphones, operating systems, and modern browsers (e.g. iOS, Android, Chrome, Safari). Only the speaker needs to log in. Participants simply open a link via a generic QR code (with the standard QR code, the same code for everyone in the room). No app, no hardware, no installation.

No. A stable internet connection is required for both the guide (speaker) and participants (listeners), since the system operates in real time and is cloud-based.

The required bandwidth depends on the number of simultaneous participants. As a guideline: up to 50 participants: approx. 25 Mbps, 50–200 participants: approx. 100 Mbps, more than 200 participants: approx. 500 Mbps download bandwidth. Each listener typically requires about 40 kbit/s to receive the audio translation (for some languages up to 128 kbit/s). The speaker's device requires approximately 256 kbit/s upload bandwidth. For stable operation, we recommend adding a safety margin of about 20%.

No. Nubart TRANSLATE works over any internet connection (Wi-Fi or mobile data) and does not require a separate network or local infrastructure. Participants do not need access to internal company networks. For very large events, a dedicated Wi-Fi network can be beneficial to optimize available bandwidth, but it is not a requirement.

There is no technical limit. Hundreds, or even thousands, of listeners can join a session simultaneously using the same standard access QR-code.

What's best: Nubart TRANSLATE will not charge you based on the number of users!

Yes, they can submit questions in writing. These are immediately sent to the speaker's smartphone or laptop and translated into their language. The speaker receives a visual notification each time a question arrives. Questions are stored until the speaker decides to answer them.

Nubart TRANSLATE currently supports Arabic, Bulgarian, Catalan, Chinese (Cantonese and Simplified Mandarin), Croatian, Czech, Danish, Dutch, English, Estonian, Filipino/Tagalog, Finnish, French, German (Germany and Switzerland), Greek, Hebrew, Hungarian, Indonesian, Italian, Japanese, Korean, Latvian, Lithuanian, Norwegian, Persian, Polish, Portuguese (European and Brazilian), Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swedish, Thai, Turkish, Ukrainian, and Vietnamese.

Yes. Each participant selects their preferred language at the beginning of the session and can change it at any time.

Nubart TRANSLATE delivers professional-grade simultaneous translation with accuracy comparable to experienced human interpreters, depending on audio quality, speaker clarity, and topic complexity.

As with human interpretation, factors like background noise, speaker pace, accents, and highly specialized terminology can affect results. To optimize terminology for specialized topics, we prepare the AI from your own material, such as agendas, technical manuals, speaker biographies, or presentation slides, much like briefing a professional interpreter before an assignment. You do not need to create a glossary yourself: sending us your documents is enough. See how the preparation works.

The key difference? You get professional interpretation quality at a fraction of the cost, with simple logistics and no booking constraints.

Speech recognition, translation and the synthetic voice are separate systems, and each one needs a different kind of preparation. Working from your own material, our team prepares them for every language your speakers will use on stage.

For the translation engine, we create a context text in the speaker's language, following the engine's own guidelines. For the speech recognition, we build a custom vocabulary of the names, acronyms and coined terms it would otherwise mishear. Where needed, we also set by hand what only a human can decide: terms that must never be translated, fixed translations for specific terms in each listener language, and pronunciation for the spoken translation.

AI produces the first draft, and our team reviews and refines it. Choosing what to leave out matters as much as choosing what to include, because an overloaded preparation makes results worse. This is the work of a professional briefing, not an automatic upload of your documents.

Send us whatever material you would give a human interpreter to prepare for a session: a programme, slides, a press kit, product sheets, speaker biographies, term lists. Anything that names the people, places, products and topics that will be mentioned is useful.

Upload your files in the Assets section of your password-protected customer area, which keeps your documents private. Give each file a descriptive name, for example "Presentation for event on 5 June, Room 2". Uploaded files stay in your account and can be reused for later events. Short lists of names or terms can also be sent by email.

Files must contain text (PDF, Word, PowerPoint, plain text). Scanned brochures, photos, audio and video cannot be read. Please also tell us which languages your speakers will use on stage. The languages your listeners choose do not matter for this step.

Preparation starts once your order is in place, or once you book a prepared test. If your material changes, send us the update and we adjust the preparation.

Please send everything as early as possible. We cannot guarantee that material received less than 12 hours before the start of the event is included. For events that start on a weekend or a public holiday, or on the morning after one, the limit is 36 hours (local time of the event).

No. Preparation makes correct results more likely, but it does not replace good audio. Background noise, speaker pace and accents affect what the speech recognition hears, and a name spoken differently from how it is written can still be misunderstood.

Preparation is also selective by design. Putting everything from your documents into the system would make results worse, so we choose what each system really needs: names, acronyms, coined terms, unusual spellings and ambiguous words. And the system can only work with what your material contains.

If a term went wrong at your event, tell us. We can add it to the preparation for the next one.

No. The free trial uses the standard system, without event-specific preparation. We prepare the AI for your event once your order is in place.

If you want to test the system with a prepared AI before ordering, we offer 30 additional minutes with prepared AI and your logo and colors for €60 (excl. VAT). Upload your material, as you would for a real event, and we prepare the system for your test.

The translation itself typically takes less than a second.

Any noticeable delay is usually linguistic rather than technical: the system waits until the speaker completes a meaningful unit of speech. In languages like English or Spanish, this can be a sentence fragment, while in verb-final languages such as German, Japanese, or Korean, a full sentence is often required.

A small additional delay comes from text-to-speech. In most languages this is barely noticeable, but for less common languages it may take slightly longer.

If the speaker talks faster than the generated voice, the translated audio may lag behind simply because speech takes time to deliver.

Maximum data security and strict GDPR compliance: Unlike many standard AI tools that use event data to train their models without asking, your content is completely safe with Nubart.

Where is your data processed? The main languages (such as English, German, French, Spanish and many more) are processed entirely on secured AWS servers within the European Union.

For some rarer world languages, we additionally use the technology of a highly specialized partner for AI voice synthesis. This component is activated only when a listener at your event selects such a language. In these cases, too, your data remains protected: the data transfer takes place exclusively on the basis of the EU Standard Contractual Clauses (SCCs).

Our security guarantees to you:

• No AI training: All partners are strictly bound by contract never to use your data to train AI models.
• Data minimization: No audio recordings of the original voices, no metadata and no personal data are transferred to the USA – only the plain, translated text for audio conversion.
• Flexible, tailored deletion periods (you decide): Data transfer is always encrypted via TLS 1.3. You decide how long we keep your transcripts and translations when you place your event order:
  • Maximum security: Immediate deletion right after processing.
  • Transcript retrieval: Deletion after 30 days, so you can export your records at your leisure.
  • Permanent storage: Guaranteed without AI training here, too, of course.

No. Our pricing is based only on the time you actually use. When the speaker mutes the microphone or closes the tab, the usage timer stops immediately (the counter will turn black so you know it’s not counting). Minutes when you are not actively speaking are never billed. And if the speaker accidentally leaves the tab open and forgets to mute it, our system will automatically stop counting after one hour of silence. This ensures you pay only for the true translation time you use—never for gaps, silences, or technical oversights.

Each speaker or room using live translation at the same time counts as one active stream. For example, if 3 speakers are talking in 3 different rooms for 1 hour, that's 3 hours in total.

Yes. Although Nubart TRANSLATE was designed for face-to-face events, it can also be used online. The speaker logs into Nubart in one browser tab and joins Zoom, Google Meet, or a similar platform in another. The meeting platform allows participants to see who is speaking and to share screens, while Nubart provides the translation. The audience receives a QR code or link beforehand, so they can follow the meeting video in one tab (or device) and listen to the translation in another.

For hybrid events where a remote speaker addresses an in-person audience, we recommend routing the videoconference audio directly into a device running Nubart TRANSLATE using a 3.5mm TRRS audio cable (male-to-male, with 3 insulating rings—TRS cables with only 2 rings won't work). This ensures clean audio input without room interference. Note that standard iPhone adapters don't support TRRS; if using an iPhone, you'll need a Lightning-to-3.5mm adapter explicitly designed for 4-conductor cables. We strongly recommend testing this setup several days before your event.

Participants can still ask written questions. These are automatically translated into the speaker's language and appear directly in the Nubart interface—either on the browser tab or on a separate mobile device, depending on how the speaker chooses to connect.

For multilingual discussions, organizers can give all participants "speaker" access in Nubart TRANSLATE. The QR code will then reproduce what each participant says, while the muted meeting platform shows who is speaking.

While some meeting platforms now offer automatic translation features, Nubart TRANSLATE stands out for organizations requiring higher confidentiality and broader language coverage. Because microphones in Zoom, Meet, or similar platforms remain muted, all speech is processed exclusively through secure, GDPR-compliant providers selected by Nubart. Meeting content is never used for external AI training or advertising and can be permanently deleted upon request after the event. With this approach, 35+ languages can be translated securely, giving you flexibility and peace of mind.

Most AI interpretation platforms rely on a generic QR code or shared link for listener access — which functions as an open URL that can be forwarded to anyone. A passcode adds little protection, since it can be shared just as easily.

For confidential events — board meetings, investor presentations, internal strategy sessions — Nubart TRANSLATE offers LWAC-protected access cards, a technology internationally protected with granted patents. LWAC binds the QR token to the first device that activates it. If a listener attempts to share their active access link, the system detects the device mismatch and denies access. No registration or personal data is required from attendees.

For public or non-sensitive events, a standard QR code remains the simplest and most practical option. Read more about listener access security in AI interpretation →

Curious how Nubart TRANSLATE compares to other providers beyond security? See the full comparison →