Live spoken translation for client meetings. Your client speaks Italian; seconds later you hear them in English, in a voice that can sound like theirs.
3 questions · 1 minute · no sign-up · you get a pilot brief
Your colleague should be contributing to the meeting, not spending it translating for everyone else.
The story
How it works
The client talks in Italian, as usual. VocalSpan listens through your microphone.
Speech becomes text and the text is translated phrase by phrase. In earlier recorded-input tests, the first English arrived after a few seconds. Live delay varies.
Spoken English in your headphones, optionally in a voice matched to the speaker, only with their consent.
Is your meeting a fit? Three questions, and you get a pilot brief to share with your team.
Where you use it
Follow a Google Meet call in your language through headphones. Meeting-audio routing is still being built; the prototype uses your microphone today.
One laptop on the table captures the speaker; the person who needs translation listens on headphones. Room acoustics and cross-talk are still being tested.
Everyone joins one session and picks a listening language. Multi-participant routing is on the roadmap, not in today’s prototype.
Earlier-engine test results · 27 September 2026
From speech start, in 14 recorded-input streaming simulations. Median lag 3.8–5.6 s by clip; worst 9.6 s. Cold start excluded.
Three podcast clips (book passages: 11.3–13.7%). Measures transcription, not translation quality.
Designed to approximate the speaker’s vocal character, only with consent. Quality varies. A neutral voice is used until matching is ready.
Internal tests, 27 September 2026: Nemotron-320 → Qwen translation → Qwen speech on an L4 GPU, recorded Italian speech replayed at real-time pace, 14 clips. These figures describe the earlier Nemotron pipeline, not the current Soniox configuration. They are development results, not independent benchmarks or live-meeting guarantees. A cloud smoke test produced first English audio 4.6 s after speech began, after a 48 s cold connection.
Languages
Every other pair gets tested before we offer it.
The film
Our first brand story is Babel, told in torn paper: people could not build together because they could not understand each other. The full film is in production; here is a first look.
Preview · 0:05 · no audio · illustrative, not a translation recording
Pricing
Final pricing is being set with our first pilots.
Guided pilot
$149 / 30 days
One recurring meeting, set up with you.
Proposed USD offer, available after setup validation. Taxes and any extra usage must be quoted before purchase. No checkout or charges on this site.
Team
After the pilot
Planned features for teams with several multilingual clients. Not an available subscription.
Organisation
Custom
Proposed scope for larger firms. Processing terms, security review and support are not yet agreed.
FAQ
A browser prototype: microphone in, spoken English out, starting with Italian → English. It is early software and may take about a minute to warm up.
Not yet. Meet is a planned workflow. Today the prototype listens through your microphone, so we check your meeting setup before any pilot.
Only if the speaker agrees. Voice matching is optional; a neutral voice is used until it is ready. Only use voice matching with the speaker’s informed permission.
In the cloud, not on your device. Confirm processing location and retention with us before using confidential conversations.
No. Expect the first English a few seconds after the speaker starts. Quality depends on language, audio and the conversation.
Pick your language pair, meeting setup and how often you meet. You get an assessment and a downloadable brief. Nothing is sent to a server.