Resources

Questions we get asked

The things event organisers and AV teams ask before booking. If yours is not here, ask us.

Does our audio go to the cloud?

No. Everything runs on a machine we bring into your venue. There is no API key and no upload, and once set up it works with the internet disconnected.

How fast are the captions?

Roughly one to three seconds after the words are spoken. Part of that delay is deliberate: it is the time spent confirming a word is right rather than showing a guess and correcting it in front of the room.

How accurate is it?

It depends heavily on the language and on your audio. We do not publish a single accuracy figure, because one number across two hundred languages would be meaningless. We would rather run a test with your material before you commit.

Do attendees need an app?

No. They scan a code and a page opens in their browser. They pick a language and can change it at any time, mid-sentence if they like.

What if attendees are not on the venue WiFi?

Guest networks often block devices from reaching each other, and plenty of people stay on cellular. For those events we publish caption text through our server so any phone can read along. It is off by default and we agree it with you first. Audio never leaves the venue either way.

Can we use a laptop microphone?

For a small meeting, sometimes. For a real event, no. A microphone in a large room picks up reverb, ventilation and the audience, and nothing downstream recovers from that. We take a line from your sound desk.

How many languages at once?

It depends on the hardware we bring, which we size to the event. The system measures the machine and sets its own limit rather than guessing. Tell us what you need and we will tell you what it takes.

Can it translate speech into speech?

Yes, for about sixteen languages. Attendees choose audio instead of captions and listen on their own headphones. The voices are synthetic and sound it. They are clear and easy to follow, but nobody will mistake them for a person. Captions cover far more languages than audio does.

Do we get a transcript?

Yes, in every language that ran, as SRT, VTT, plain text or JSON. It is written as the event runs, so it survives the machine being switched off mid-session.

Is this as good as a human interpreter?

No. For negotiation, legal testimony or anything where a misunderstanding is expensive, hire one. SpeakMany is for the far more common case where the alternative is no translation at all.

What happens if something goes wrong during the event?

An operator is there. The console reports faults in plain language (low input, audio being discarded, a misconfigured pipeline) rather than failing quietly, and the transcript keeps writing throughout, so nothing already said is lost.

What do you need from us?

A feed from the sound desk, access to set up and test beforehand, and a list of names, products and acronyms so we can load the glossary. See terms for the full list.