Live transcription, and the minutes on your desk the next morning
The same thing that puts text on screen while somebody speaks leaves you the full transcript when they stop. Not a second service: the same session.
- transcripts kept from the live path
- 0transcripts kept from the live path
- and the minutes self-delete
- 30 daysand the minutes self-delete
- and VTT for the video
- SRTand VTT for the video
- regions to transcribe in, your choice
- 3regions to transcribe in, your choice
What is live transcription?
Turning speech into text as it is spoken, rather than recording first and transcribing later. The practical difference is not the speed: it is that the text is useful during the event (it can be projected, translated and read on a phone) and not only afterwards.
The problem with minutes
- Somebody takes them by hand, so they arrive days later and contain only what that person judged important.
- Automatic transcription done afterwards has no real clock times, so nothing can be cited as "what was said at 11:40".
- With several speakers the system guesses who spoke, and attributing a decision wrongly is worse than not attributing it.
- And audience questions and votes appear nowhere at all.
What you have when it ends
The full transcript with real clock times
Plain text stamped with the local time each line was spoken, not elapsed time since the start.
SRT and VTT subtitles
With times relative to the start, which is what the session video needs.
AI minutes, in the language you choose
Topics, decisions, action items with an owner and a due date, figures, disagreements and quotes.
Votes and questions, inside the minutes
They enter the document as recorded data, not as a paraphrase.
Each speaker’s name
With two or more microphones the technician types them. With one, nothing is attributed, on purpose.
What happens during the event
The text goes on screen while people speak
Over NDI, over HDMI, on any TV in the venue or on the attendee’s phone.
Translated, if you need it
Up to 12 languages at once from a catalogue of 58, chosen by each person on their own phone.
The operator sees what is being transcribed
With low-confidence words flagged.
Custom vocabulary, loaded beforehand
Names, brands and acronyms set in advance. It is the single biggest lever on accuracy.
What is stored and what is not
We store no transcript of the live event
The live path writes it to no database of ours. Our primary translation provider does hold its API requests for 30 days for abuse monitoring: it is in the privacy policy.
The one exception is what you choose to keep
The AI minutes, or the transcript kept when closing the app. Only if you ask, you are told on screen, and it is deleted after 30 days.
Audio is transcribed in your account’s region
Europe, the Americas or Asia-Pacific: set from your country, changeable by the account owner. Every request we send to Deepgram opts out of retention and model training, and the fallback runs in the same region.
The sub-processor list is published
Who each one is, what they handle and where. On the website, not in a PDF on request.
Accessibility
Transcription and AI minutes on one microphone. The AI minutes are included in every paid plan and are never metered separately.
Frequently asked questions
Are the minutes written by a person?
No, a model generates them from the transcript against a fixed shape: decisions, action items with an owner and a due date, figures, disagreements and open questions. You can regenerate them in another language.
Does it tell speakers apart?
Only when each person has their own microphone, and with the name the technician types. With one mixed microphone nothing is attributed: guessing who said what produces minutes that blame the wrong person.
What language are the minutes in?
Whichever of the 58 catalogue languages you pick, regardless of the language that was spoken.
How long is the transcript kept?
30 days, and only if you asked for the AI minutes or kept it when closing the app. After that it is deleted automatically.
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.