Video, transcription, translation, AI reports and voice trust — natively integrated, sovereign by design. No add-ons. No US intermediaries.
Unlimited HD meetings, breakout rooms, whiteboard, screen sharing and moderated meetings. Enterprise-grade security is the default, not an upgrade.
HD meeting room · breakout · whiteboard · screen share
Live, searchable transcription with 95%+ accuracy — every participant follows the meeting as text, in their own language, in real time. Speech models are hosted and run by us on EU infrastructure — on GPU in our data centres, and on CPU or ARM with no GPU card for on-premises and embedded deployments — so your audio never touches an external API.
Live transcript · speaker labels · language per participant
The moment a meeting ends: executive summary, decisions, action items with owners, and audit-ready metadata. Processed in RAM-only secure enclaves in the EU and wiped after the session. Your content never trains anyone's model.
Summary · decisions · action items · exportable report
Instant translation of the meeting feed into 47 languages and growing, in under 300 ms — practically instantaneous, so the conversation never waits for the subtitles. The engine learns your company's own vocabulary, and it runs on open-source libraries we host ourselves: sovereign end to end.
Roadmap: the translated feed delivered as speech, in the speaker's own voice — live today in Voice Trust+, see below.
47 languages · custom vocabulary · translated minutes
Real-time spectral analysis of every voice in the room. A passive green "Verified Voice" shield per participant; amber and red alerts on anomaly; host actions (verify identity, mute, report) and a 2FA identity handshake for flagged participants.
Privacy-preserving by design: no permanent voice templates (GDPR Art. 9), risk scores rather than automated decisions (GDPR Art. 22), transient data deleted at meeting end.
Illustration of the Verified Voice interface. The percentage shown is per-speaker confidence on a single alert — distinct from CandyVoice's 99.1% detection accuracy (99.4% of real voices recognised as human, 3% EER under noise) (internal benchmark; independent validation planned). Deepfake defence is live in production, on by default on Pro Security.
Speak in your own language; every participant hears you in theirs — in your voice, with your intonation, across 47 languages. Not a synthetic narrator reading out a translation: your own voiceprint carries the message, so emphasis, hesitation and tone survive the language barrier. That is what makes a negotiation or a board call in five languages feel like one conversation instead of five.
The same module runs in reverse. Voice anonymisation replaces your voiceprint in real time, so what leaves the room cannot be captured, stored or cloned — for whistleblowing hotlines, sensitive negotiations and protected witnesses. Privacy-preserving by design: no permanent voice templates (GDPR Art. 9), transient data deleted at meeting end.
Both ship as the Voice Trust+ add-on (+€4.99/user/month on Starter or Pro Security, available today), Article 50-compliant by design — explicit speaker consent and machine-readable marking of synthetic audio — alongside studio-grade noise filtering, from the same CandyVoice audio module.
Your voice across 47 languages · anonymised when it matters
A 30-minute private demo on your use case: real-time transcription, AI minutes, live translation and the Verified Voice shield.
Book a private demo See pricing