Skip to main content
Real-time voice moderation stream (WebSocket)
curl --request GET \
  --url wss://voice.moderationapi.com/v1/stream \
  --header 'Authorization: <authorization>' \
  --header 'Content-Type: application/json' \
  --header 'Sec-WebSocket-Protocol: <sec-websocket-protocol>' \
  --data '
{
  "event": "start",
  "streamSid": "<string>",
  "mediaFormat": {
    "encoding": "audio/x-mulaw",
    "sampleRate": 8000
  },
  "tracks": [
    {
      "authorId": "<string>"
    }
  ],
  "conversationId": "<string>",
  "channel": "<string>",
  "emitPartials": false,
  "metadata": {}
}
'
{
  "v": 1,
  "event": "session.started",
  "sessionId": "<string>",
  "tracks": [
    "<string>"
  ],
  "conversationId": "<string>"
}
Moderate live voice and call audio over a WebSocket. You send a start frame, stream media frames as audio arrives, then stop; the server transcribes speech and returns a moderation verdict for each finalized utterance. For the full walkthrough and code examples, see Real-time voice.

Headers

​
Authorization
string
required

Bearer <api_key>

​
Sec-WebSocket-Protocol
enum<string>
required

Requested subprotocol.

Available options:
moderationapi.v1

Body

application/json

Frames sent by the client over the socket (not an HTTP body).

First frame the client sends. Declares the conversation, audio format, and tracks.

​
event
enum<string>
required
Available options:
start
​
streamSid
string
required

Your identifier for this stream.

​
mediaFormat
object
required
​
tracks
object[]
required

One or both tracks. Stream only the track(s) you have.

​
conversationId
string

Your external conversation id. Omit to have one generated and returned in session.started.

​
channel
string

Optional. Selects which channel's policy configuration applies.

​
emitPartials
boolean
default:false

Set true to also receive interim, non-final transcripts.

​
metadata
object

Arbitrary JSON attached to the conversation. Stored as-is and not interpreted by moderation.

Response

101 - application/json

Switching Protocols. The server then streams event frames over the socket; the key one is utterance.final.

Sent after the start frame is accepted.

​
v
enum<integer>
required
Available options:
1
​
event
enum<string>
required
Available options:
session.started
​
sessionId
string
required
​
tracks
string[]
required
​
conversationId
string