Every screen and control named on this page is checked against the product at build time. If a name here does not match your screen, your workspace is on an older release.

Which speech engine turns your AI notes into text

Before the AI can write a note, the sound of the conversation has to become words. SISC tries the speech engines below in this order and uses the first one that can listen. Whichever it is, SISC never keeps the sound — only the words.

You need

• ai.notes.create

Steps

1. On your computer, always tried first. The speech model runs inside your browser, so the sound never leaves your computer and nothing is paid per minute. It needs a modern computer — a graphics card the browser can use, or plenty of memory — and the model is downloaded once from SISC, the first time; after that it starts straight away.
2. Your company’s own speech server. If your company runs one and SISC support has connected it, the sound goes there and to nobody else. A server that tells voices apart also marks who spoke.
3. SISC AI’s speech service, used next. It tells voices apart and keeps the same person as the same speaker for the whole recording, and its copy of the sound is deleted as soon as the words come back.
4. SISC AI’s backup speech service, used last. It writes the words but does not say who spoke.
5. To see which engine heard a note, open the note and look under How it was heard: it names the engine, and says Speakers: yes when the words are marked by speaker.
6. When it does, write each person’s name under Who was speaking? and press Save names. If one person was heard as two speakers, give both the same name.

Speech is turned into text on your computer first; only if that isn’t possible is it sent to the AI service chosen by your company, and the sound is never kept. If no engine can listen, SISC says so before you start rather than after.

Related

• Take AI notes of a meeting or a call
• Take AI notes from the AI box, with or without SISC Boardroom