Meeting bot glossary
Short, practical definitions for the terms you will meet when building on meeting bot infrastructure. Each entry links to a deeper explainer with examples and API context.
Meeting bot
A software participant that joins a video call to capture audio, video, transcripts, and events on your behalf.
Meeting bot API
The programmatic interface for sending bots into meetings and getting structured meeting data back.
AI notetaker
An application that records a meeting and turns it into summaries, action items, and searchable notes.
Speaker diarization
The process of splitting a transcript by who said what, so every line is attributed to a speaker.
Real-time transcription
Speech converted to text while the meeting is still in progress, typically delivered over a streaming connection.
Per-participant audio
Separate audio streams for each person in a meeting, instead of one mixed track.
Meeting intelligence
The broad category of products that extract insight from meetings, including coaching, analytics, and CRM enrichment.
Webhook
An HTTP callback that notifies your application when something happens, like a recording becoming ready.
WebSocket streaming
A persistent two-way connection used to push live meeting audio, video, and transcript events to your app with low latency.
Calendar integration
Connecting calendars so bots schedule themselves into upcoming meetings automatically.
Frequently asked questions
What is Zoom RTMS (Realtime Media Streams)?
Zoom RTMS is a native pipeline that gives your app live access to a meeting's audio, video, and transcript over a WebSocket, including per-participant streams. It is Zoom's own vendor-layer approach, so instead of a client bot you subscribe to the media Zoom emits. It covers Zoom only, so if you also need Google Meet and Microsoft Teams you either run RTMS alongside separate integrations or use a bot-based API like MeetStream that deploys your own bot into all three platforms behind one interface.
What is speaker diarization and why does a meeting API need it?
Speaker diarization answers who spoke when, segmenting an audio stream into speaker-attributed turns so a transcript reads as a conversation rather than one block of text. Diarization is a prediction, so accuracy drops with cross-talk and similar voices. A cleaner path is per-participant audio: when each speaker arrives on their own isolated stream keyed by participant ID, attribution is already exact and you can skip the diarization guesswork.
What is per-participant audio and how is it different from a mixed recording?
A mixed recording blends every speaker into a single track, while per-participant audio delivers each person on a separate stream tagged with their participant ID. That separation lets you run transcription, sentiment, or coaching logic on one speaker without bleed from the others, and you get correct attribution without relying on diarization. MeetStream delivers these isolated streams over WebSocket in real time at roughly 200ms across Zoom, Google Meet, and Teams.
What is a meeting webhook and how does it work?
A meeting webhook is an HTTP callback that pushes an event to your endpoint the moment something happens, so you are not polling to ask whether a bot joined or a recording finished. You register a URL, and the service POSTs a JSON payload on events like bot joined, participant joined, transcript ready, or meeting ended. With MeetStream you point webhooks at your own endpoint and receive participant events and status changes in real time, which is what lets you drive automations like CRM updates without a polling loop.
From definitions to production
Everything on this page ships in one API. Join, record, stream, and transcribe meetings programmatically.