Sending Oracle Radio WebSocket Audio

Send each audio packet as an unframed binary WebSocket message during an active Radio audio transmission.

Permissions and Prerequisites

The following permissions and prerequisites apply to this operation.

Requirement	Value
Authentication	Complete the supported authentication path for the client before sending this operation. See authentication.
Radio permission	`radio.audio.transmit`
Channel access	The authenticated user must have talk access to the active channel.
Session state	The WebSocket connection must be open, the talk token must be valid for the active channel, and the client must have sent `START`. Use only the token granted to the authenticated user, and don't transfer it to another user or client.

Request

This binary input has no JSON parameters and no per-message Content-Type field. Send one complete binary WebSocket message that contains only the encoded audio packet.

For a JavaScript WebSocket client, call this function after successfully sending START. The encodedAudioPacket argument contains one packet encoded by using the audio settings declared in START.

function sendAudioPacket(webSocket, encodedAudioPacket) {
  if (webSocket.readyState !== WebSocket.OPEN) {
    throw new Error("WebSocket connection is not open");
  }
  webSocket.send(encodedAudioPacket);
}

Payload

The following table describes the binary audio payload sent by the client.

Attribute	Value	Description
WebSocket message type	Binary	Use a binary WebSocket message, not a JSON text message.
API content type	`application/octet-stream`	This is the media type in the API contract. It isn't a field or header in the WebSocket message.
Data	Encoded audio bytes	Send one audio packet in the format declared by `START`. Don't add the channel prefix used by server-to-client audio.

Protocol Constraints

Follow these requirements when sending audio packets during an active transmission.

Constraint	Requirement
Transmission state	Send packets after `START` and before `END`, while the talk token is valid.
Packet encoding	Use the format, sample rate, channel count, bitrate, and frame size declared by `START`.
Audio-session size	Send at most 50 MiB of audio for one session.
Metadata packets	The service recognizes packets starting with `OpusHead` or `OpusTags` as metadata and ignores them.

Responses

The following table lists the messages the client can receive and explains their purpose.

Message type	Kind	Description
Framed audio packet	Binary relay	Other subscribed connections receive the original audio packet with a channel-name prefix added by the server.
`TOKEN_INVALID`	Direct error	Indicates that no current talk token controls the transmission. Stop sending audio.
`ERROR`	Direct error	Indicates that the audio session exceeded 50 MiB or another packet-processing error occurred.
`INTERRUPTED`	Asynchronous event	Indicates that a higher-priority user took the talk floor. Stop sending audio.
`TOKEN_EXPIRED`	Asynchronous event	Indicates that the talk token expired. Stop sending audio.
`TRANSMISSION_TIMEOUT`	Asynchronous event	Indicates that the transmission timed out and the token was automatically released.

Note

The initiating session doesn't receive a relayed copy of its binary audio. Continue based on local WebSocket send completion; don't wait for an audio echo.

Framed Audio Packet

The following table describes the parts of an audio packet relayed by the server.

Part	Data type	Required	Description
Channel-name length	2-byte unsigned integer	Yes	Length of the UTF-8 channel name in big-endian byte order.
Channel name	UTF-8 bytes	Yes	Source channel. Its length is specified by the first two bytes.
Audio packet	binary	Yes	All remaining bytes. These are the original client audio packet.

Note

Client-to-server and server-to-client binary messages use different layouts. Don't send the server-added channel prefix back to the service.

Oracle Cloud Infrastructure Documentation