• command (string, required)

• text (string, required) Plain text or SSML to synthesize and play on the active voice channel.

When streaming, send this field in multiple messages with isLast: false until the final chunk, which must have isLast: true.

When using SSML, tags must not span across message boundaries within a batch. Each chunk must be independently valid XML if SSML is used.

• format (string) Enum: TEXT, SSML

• ttsVoice (string) Name of the TTS voice to use for this batch.

Only the value from the first message in a batch takes effect; this field is ignored on subsequent messages in the same batch. If omitted, the default voice configured when the session was initiated is used.

• isLast (boolean, required) Indicates whether this is the final message in the current TTS batch.

Set to false on all intermediate chunks and true on the last chunk to signal the end of the batch and trigger an audio flush. The next message with isLast: false begins a new batch.

• isInterruptible (boolean) Controls whether speech detected on the voice channel can automatically interrupt this TTS batch.

Overrides the enableInterruptions setting from the session for this batch only. Even when set to false, playback can still be stopped by explicitly sending an interrupt command over the WebSocket. Defaults to true when not specified.