Simultaneous translation moves into the control room.

VoxBit is the control-room appliance that translates the speaker in real time: synthetic voice to headphones, live subtitles, RTMP streaming and an archive master — from a single unit, inside your signal flow. It connects to the desk like any outboard processor. Any AV technician can run it, from a tablet.

Buy hours Contact us for a demo

LATENCY ~2S · 50+ SOURCE LANGUAGES · 1 TARGET LANGUAGE / SESSION · RTMP · MKV MASTER

02 / THE CHALLENGE

Traditional simultaneous interpreting doesn't scale.

Anyone who runs multilingual events knows it: traditional simultaneous interpreting is the budget line that never scales. And when the budget runs out, translation is simply cut.

thousands of €

PER DAY · PER LANGUAGE

Indicative market costs for a professional interpreting shift: soundproof booth, a pair of interpreters, dedicated distribution system. For a single language.

× N languages

THE MULTIPLIER

Every additional language multiplies interpreters, booths and crew. Costs grow linearly, budgets don't.

0 archive

WHEN THE EVENT ENDS

The translation evaporates when the lights go down: no transcript, no reusable master, nothing to hand over to the client.

For high-stakes settings, human interpreters remain the standard. VoxBit serves the events where translation simply doesn't happen today.

03 / THE SOLUTION

One unit in the control room, five modules at work.

TRAD

Real-time voice translation

~2 s latency, 50+ source languages with auto-detect, a constant, recognisable synthetic voice. Transparent session-swap covers 8+ hour events without interruption.

Media processing is local; the AI model is cloud-based (OpenAI).

SUBS

Live multi-target subtitles

An HTML page served by the appliance for OBS, vMix or the venue projector. Optional burn-in on the on-air feed, live toggle without stopping the translation.

RTMP

Streaming to the platforms

NVENC H.264 hardware encoding at 1080p to Vimeo Live, YouTube, Twitch. Audio your way: floor, translated or mix.

REC

MKV archive master

Local pre-overlay recording, in parallel with the stream. The master stays on the appliance, ready for post-production.

POST

Post-event analysis and bundle

Summary, keywords, extracted Q&A, back-translation into further languages. Branded PDFs and a ZIP bundle with an offline viewer to hand to the client.

04 / HOW IT WORKS

Inside your signal flow, like outboard gear.

IN 1 · AUDIO

HOUSE DESK

post-fader feed → USB audio interface

IN 2 · VIDEO

HDMI

speaker / slides → USB capture

OPENAI REALTIME · TLS 1.3

audio only, encrypted — media processing stays on the machine

VoxBit

appliance in the control room

OUT 1~2 S

RETURN TO DESK

translated headphone bus

OUT 2

RTMP

Vimeo Live / YouTube / Twitch — optional burn-in subtitles

OUT 3

REC

local MKV master

OUT 4

SUBTITLE PAGE

OBS / vMix / venue projector

▪ BYPASS — original audio on air, zero latency, independent of the network

IN 1 · AUDIO

HOUSE DESK

post-fader feed → USB audio interface

IN 2 · VIDEO

HDMI

speaker / slides → USB capture

VoxBit

appliance in the control room

OPENAI REALTIME · TLS 1.3

audio only, encrypted — media processing stays on the machine

OUT 1~2 S

RETURN TO DESK

translated headphone bus

OUT 2

RTMP

Vimeo Live / YouTube / Twitch — optional burn-in subtitles

OUT 3

REC

local MKV master

OUT 4

SUBTITLE PAGE

OBS / vMix / venue projector

BYPASS — original audio on air, zero latency, independent of the network

The signal comes in from the house desk and HDMI, VoxBit translates it via the cloud model, and returns translated headphones, RTMP streaming, an MKV master and subtitles — with a zero-latency BYPASS always armed.

BANDWIDTH 5–15 MBPS SYMMETRICRTT < 100 MS OPTIMALUSB CLASS-COMPLIANT AUDIONO INBOUND NATETHERNET OR WI-FI

Typical setup: 15 minutes. Audio in/out, HDMI, network, event profile, live.

05 / LISTENING TEST

HEAR ME SPEAK

Before you take our word for it, listen.

In English

Italian speaker → VoxBit translates to English In Italian

English speaker → VoxBit translates to Italian

Scan the QR with your phone and hear the translation live.

06 / USE CASES

Three setups, three different control rooms.

SETUP · 1× VOXBIT · IT → EN · HEADPHONES + YOUTUBE

International medical congress

EXAMPLE SCENARIO

INPUT: House desk, IT floor
TARGET LANGUAGE: EN
OUTPUT: Headphone bus + YouTube Live
DELIVERABLE: Transcript + MKV master

500 attendees in the hall, Italian floor. The translated voice reaches the international audience's headphones and the English YouTube stream, with live subtitles.

One target language per session: for EN+ES+FR you run three appliances side by side. And for high-stakes clinical sessions, human interpreters remain the standard.

RESULT · EN ON HEADPHONES AND STREAM, NO BOOTHS

SETUP · 1× VOXBIT · IT → EN · RTMP MIX

Corporate webinar

EXAMPLE SCENARIO

INPUT: Speaker audio via desk
TARGET LANGUAGE: EN
OUTPUT: RTMP, floor/translated mix
DELIVERABLE: Transcript + same-day summary

Speaker in Italy, global audience. The stream goes out as a floor/translated mix; at the end of the session the transcript and executive summary are ready to distribute the same day.

RESULT · DELIVERABLES SHIPPED THE SAME DAY

SETUP · 1× VOXBIT · IT → EN · VIMEO LIVE + REC

Broadcast festival

EXAMPLE SCENARIO

INPUT: Video control: HDMI + audio
TARGET LANGUAGE: EN
OUTPUT: Vimeo Live, 1080p burn-in
DELIVERABLE: MKV master + ZIP bundle

Vimeo Live streaming with burned-in subtitles, NVENC 1080p encoding. In parallel, the pre-overlay MKV master stays on the appliance for post-production, and the ZIP bundle with offline viewer goes to the client.

RESULT · MASTER TO POST, BUNDLE TO THE CLIENT

07 / PRICING

ON REQUEST

Hour packs, no forced subscription.

VoxBit runs on hour packs: you pay for the translation time you use. The positioning is a low-overhead AI service — a reasonable markup on the model's real costs, not American-style premium pricing.

The full price list is available in the reserved area after registration.

08 / FAQ

The questions ICT asks (and the director too).

Does our audio end up in OpenAI's training data?

No. OpenAI works on an opt-in model: no API data enters training unless the customer explicitly chooses to share it — and VoxBit never enables that option. Payloads are retained for up to 30 days for abuse monitoring, then deleted. For strict policies, a Zero Data Retention configuration is available on request via the OpenAI Trust Portal.

OPT-IN NEVER ENABLED · RETENTION MAX 30 DAYS · ZDR ON REQUEST

How much latency does the translation have?

Around 2 seconds, from the start of the sentence to the voice in the headphones. It isn't a technical limit but a quality choice: the model could reply even faster, but — exactly like a professional interpreter — it takes a moment to grasp the structure and context of the sentence before rendering it. Translating word by word on the instant would multiply errors: those ~2 seconds are the time to understand, not a lag you're stuck with. Imperceptible in conferences, keynotes and panels; noticeable only in the fastest back-and-forth.

LATENCY ~2 S · THE TIME TO UNDERSTAND, LIKE AN INTERPRETER

Does it work without internet?

Not for translation: it needs the real-time connection to the model. But local MKV recording keeps running offline, and BYPASS keeps the original audio on air at zero latency, independent of the network.

LOCAL REC ALWAYS ON · ZERO-LATENCY BYPASS

How many languages per session?

One target language per session. To serve more languages you run multiple appliances in parallel. Simultaneous multi-language on a single unit is on the roadmap.

1 TARGET / SESSION · N LANGUAGES = N APPLIANCES

How much bandwidth does the venue need?

2 to 5 Mbps symmetric for translation alone, 10 to 20 Mbps with 1080p streaming. The critical direction is upload. No inbound NAT from the public internet: the AI session travels outbound-only on port 443. Outbound access is also needed for DNS, NTP and — where used — the update repo (80/tcp) and plain RTMP (1935/tcp): the complete firewall rules are in the technical brief for ICT.

2–5 MBPS TRANSLATION · AI SESSION OUTBOUND 443 · NO INBOUND FROM INTERNET

Is there a DPA?

Yes. OpenAI provides its standard DPA, downloadable from the Trust Portal; MeetBit provides its own DPA to the customer during onboarding.

OPENAI DPA (TRUST PORTAL) + MEETBIT DPA

09 / CASE STUDY

Case study coming soon — we are documenting the first events in production.

Want to be the first? Let's talk →

10 / CONTACT

Let's talk about your next event

You'll get answers from people who actually run control rooms.

Write to us: support@meetbit.it Buy hours

What to tell us for a useful demo

DATE AND VENUE
FLOOR AND TARGET LANGUAGE
STREAMING YES/NO AND PLATFORM
VIDEO CONTROL ROOM ON SITE