Video and Audio

Stream Video runs video calling, audio rooms and livestreaming on Stream's global edge network. The SDKs manage devices, call state and participant tracks; the same API scales from one-to-one calls to streams with millions of viewers.

How the pieces fit together

Devices join calls through a client SDK, which handles signaling, devices and call state against Stream's edge network. Your backend uses a server-side SDK to create calls, configure call types and sign the user tokens clients join with.

flowchart LR
  subgraph app[Your app]
    client[Video client SDK]
  end
  subgraph backend[Your backend]
    server[Server-side SDK]
  end
  client <--> api[Stream API]
  server --> api
  server -. user token .-> client

SDKs

Client Side SDKs

Everything a device needs to join a call: signaling, device management and rendering, plus prebuilt call UI on the app platforms.

React
Components and hooks for web calling UIs.
iOS
SwiftUI components for calling UIs.
Android
Compose components for calling UIs.
React Native
Native calling UI from one codebase.
Flutter
Widgets for calling UIs.
JavaScript
The core browser client, with the UI left to you.
Unity
Video calls inside Unity games.
ESP32
Calling on embedded hardware.

Backend

Create calls, mint tokens and configure call types from your server. One docs set covers all seven server languages with language tabs; Vision Agents is a separate framework for putting real-time AI agents in calls.

Build with AI

Your AI agent can work with Stream from the terminal with the Stream CLI. There's also an Agent Skills pack so it can look up docs and configure your app.

curl -fsSL https://getstream.io/cli.sh | bash
getstream skills

Platform

Users, auth and the rest of the account-level machinery work the same in every Stream product, so they are documented once in the platform docs. Video apps touch these first:

Start here

Pick your platform and use case; each tutorial ends with a working app.

Video calling

Audio rooms

Livestreaming

FAQ

Are calls, audio rooms and livestreams different APIs?

No. Each is a call with a different call type; the SDKs and API are the same.

How many viewers can a livestream have?

Millions. Viewers watch over WebRTC or HLS.

Can I stream from OBS?

Yes. Livestreams accept RTMP, WHIP and SRT input.

Can calls be recorded and transcribed?

Yes. Recording and transcription are built into the API.

Do I need my own backend?

Yes. Your server signs the user tokens devices join calls with.

Can an AI agent join a call?

Yes. Vision Agents puts real-time AI agents into calls.

Can I read these docs as markdown?

Append .md to any page URL. /llms.txt indexes every page.