Verbal transcription software, while the meeting is under way

Transcribing a meeting, a conference or a public talk by hand costs hours, and in those hours pieces get lost. dotenv wrote verbal transcription software that uses machine learning to go from voice to text while the meeting is taking place.

The client is a public administration in Emilia-Romagna, and this decided the architecture before anything else. Minutes carry the names and the words of people: the audio does not leave the client’s infrastructure, the speech recognition model runs in-house, and no third-party service sits between the recording and the final text.

The software is a product of ours and can be reused: another client can have it too.

What it does while the meeting goes on

The document gets its header before the recording starts. From there the system records and converts the audio into text, tells apart the people who speak, and filters out background noise: what remains is the clean recording, without the murmur meetings always have.

The same thing works afterwards, on a recording already made.

Silences become punctuation

It is the part you notice when rereading the minutes. The software interprets pauses: a short pause becomes a full stop on the same line, a long pause a full stop and new paragraph.

There is also a spell check during dictation, which flags straight away to whoever is taking minutes the words or phrases that are unclear.

When the meeting ends

The text goes out by email to all participants, and the recording stays available to listen to again.

Around the transcription there is the management of the whole meeting (the invitations to send, the follow-ups) and the platform integrates with the CRMs already in use.

The second piece of software, born from this

For meetings, webinars and live talks dotenv then wrote real-time multilingual translation software, also based on artificial intelligence.