An NDI source with a real alpha channel, inside your signal
The main output is NDI with an alpha channel: your switcher composites it over programme like any other input. The full-screen page is there too if you want it, but you do not depend on it.
- with a real alpha channel
- NDIwith a real alpha channel
- simultaneous microphones
- 4simultaneous microphones
- to a translated caption
- ~1.7 sto a translated caption
- control with a real tally
- APIcontrol with a real tally
What usually has to be rigged
- Wiring the sound desk into a tablet through an adapter, because the system expects a consumer device.
- Leaving the captions on a separate monitor, outside programme, for want of a compositable output.
- Building a downstream chroma stage to recover transparency.
- Driving everything from a browser tab, with no physical control.
How it rigs
NDI with a genuine alpha channel
The background is transparent in the frame itself. vMix, TriCaster or OBS composite it like any other input.
Or full-screen HDMI with a chroma key
When the room is not behind a switcher.
Or a chroma-key URL in a browser
Same key colours as the HDMI output, for when no video output is left.
The engine runs on your machine
The internet carries audio and text, not your show. NDI and HDMI are drawn here.
Driven from a physical button
Bitfocus Companion, tested against the real thing, or a TCP or HTTP control API that a Crestron, AMX or Q-SYS panel can drive (not yet tested on that hardware), with a real tally.
Production control
Up to four simultaneous microphones
Each with its own colour, position and speaker name.
Captions hide without touching anything else
They still reach the audience phones.
The vocabulary is loaded before
Names, brands and acronyms pinned in advance. That is where accuracy is won.
It fails clean, never frozen
If the connection drops the output fades to transparent.
Deliverables for the end client
Video captions in SRT or VTT
Ready for next-day delivery.
AI minutes of the event
One more deliverable to invoice, with no extra work.
An audience report
How many followed, and in which languages.
Why it fits a production company
It does not change your workflow
It comes out over NDI into what you already have and adds a line to your invoice.
Driven from the surface you already use
The Companion command list is generated from the application’s own table.
Published price, direct purchase
Quote an event without waiting for a call back.
Your client preps their own event
A signed room link that opens that room and nothing else. Not your billing, not your other clients.
Let us talk about reselling
If you build congresses, write to us.
Broadcast Multi-Mic
Four simultaneous microphones, a season’s worth of hours and priority support. Event Pro if you work with two.
Frequently asked questions
Do I have to hand over my account so the client can prepare the questions?
No. There are two signed links and neither needs an account. The room link is for whoever organises the event: it opens the audience QR, the questions and votes, the AI minutes, and the results afterwards. The moderation link is narrower, for whoever holds the tablet on the day. Both work for one room only, both stop working the moment you release the code, and neither reaches your billing or your other rooms.
Is it real alpha or a chroma key?
Real alpha in the NDI frame, tested inside vMix and OBS. It is also verified against the NDI runtime on every release: the panel plate measures opaque and the rest of the frame measures transparent.
Windows or Mac?
Both, and both are signed: the Mac build carries a Developer ID signature and is notarised by Apple, and the Windows installer has carried a Certum OV certificate since 1.7.31. SmartScreen now names the publisher instead of saying "unknown", though it may still warn until the installer accrues downloads. An OV certificate gives identity, not instant reputation.
What if the venue network profile is "Public"?
Windows blocks inbound mDNS discovery for every application, not just ours. The app detects it and warns you before you go on air.
How much latency does it add?
A translated caption lands about 1.7 seconds after the word is spoken, and 2.5 seconds in the slowest one in ten. Measured, not estimated. Untranslated captions, in the spoken language, arrive sooner.
Other solutions
Try one hour free
One hour of captioning and translation on one microphone, no expiry, no card. Enough to prove it on your own rig before you pay.