Real-time translation, on the screen and in every pocket
The translation appears while somebody is speaking, not when the sentence ends. It leaves by the venue screen and by every attendee’s phone, in whichever language each person picks.
- speech to text, median, measured
- 718 msspeech to text, median, measured
- median when translated, measured
- ~1.7 smedian when translated, measured
- languages in the catalogue
- 58languages in the catalogue
- translation providers at once
- 2translation providers at once
What does "real-time" actually mean here?
That text is written on screen seconds behind the voice rather than minutes, and that it is written word by word instead of appearing all at once when a sentence closes. Measured against the deployed service, speech becomes text in 718 ms median, and translated captions land about 1.7 s after the word (2.5 s in the slowest one in ten).
The problem today
- Text appears in lumps, whole paragraphs at a time, and the room cannot read at the speaker’s pace.
- The translation falls a sentence behind and never catches up.
- Each added language is another budget line, so it gets capped at two.
- And when the venue network coughs, the last caption freezes on the big screen.
How it is made readable
It is written word by word
On a rhythm chosen around a single eye fixation, instead of dropping a whole sentence at once.
Every line is exposed for the time it takes to read
A long line stays longer than a short one, and the next one waits its turn.
Lines arrive in the order they were spoken
Each language is an independent request, so the order is rebuilt before anything is written.
Two translation providers side by side
The second starts without waiting for the first to fail.
What the attendee gets
They pick a language and change it whenever they like
Up to 12 languages at once per event, from a catalogue of 58.
They can listen instead of reading
The phone reads it aloud through their own earphones, at no extra cost.
Text is joined back into sentences on their screen
The line breaks a venue screen needs are not inflicted on somebody reading a phone.
No per-person or per-language charge
Which is what makes the third, fourth and fifth language get approved.
How it works underneath
Every line is translated in isolation
No history and no shared cache, so one event can never bleed into another’s translation.
Spoken text is data, never an instruction
If somebody says "ignore the above" out loud, that gets translated, not obeyed.
Custom vocabulary
Names, brands and acronyms loaded before the event reach the do-not-translate list, so proper nouns are not "translated".
The fallback translator runs in your account’s region
With an automated daily probe proving it is alive.
Event Pro
The first plan that translates, with all 58 catalogue languages and two microphones.
Frequently asked questions
How many languages at once?
Up to 12 reading languages at a time per event, across phones and screens, chosen from a catalogue of 58. The 58 are what an event can choose from; 12 is how many are served simultaneously.
Does each extra language cost more?
No. There is no per-language and no per-attendee charge; what is billed is on-air hours and the number of microphones in the plan.
What happens if the connection drops?
The audience output fades to transparent and reconnects by itself. A stale caption is never left frozen on screen.
Can it detect the spoken language automatically?
Yes, and if the speaker switches language mid-event the system decides line by line who needs translating and who does not.
For your kind of event
Other ways people search for this
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.