Skip to navigation
APIsInstant Voice Cloning

WebSocket

Beta
View as Markdown

Beta. WebSocket channel for real-time Instant Voice Cloning synthesis.

Protocol matches Bulbul TTS WebSocket: send config once with exactly one of a saved voice_id or base64 ref_audio (optional ref_text), then incremental text and flush messages. The server streams base64 audio frames and a final event.

Note: This API Reference page is provided for informational purposes only. The Try It playground may not provide the best experience for streaming audio. For optimal streaming performance, please use the SDK or implement your own WebSocket client.

Auth: send your key in the api-subscription-key header, or as a WebSocket subprotocol api-subscription-key.<key>.

Streaming codecs: mp3 (default), wav, linear16, mulaw, alaw. flac, aac, and opus are not supported.

Quality checks (QC/VAD) and duration bounds are not accepted on this path.

Handshake

WSS
wss://api.sarvam.ai/voices/clone/ws

Headers

Api-Subscription-KeystringRequired
API subscription key for authentication

Query parameters

send_completion_eventenumOptionalDefaults to true
Enable completion event notifications when synthesis finishes. When set to true, an event message will be sent when the final audio chunk has been generated.
Allowed values:

Send

Voice Cloning Configure ConnectionobjectRequiredBeta

Send initial configuration for voice-cloning streaming

OR
Voice Cloning Send TextobjectRequiredBeta

Send text chunk for voice-cloning speech synthesis

OR
Voice Cloning Flush SignalobjectRequiredBeta
Send signal to end text streaming for voice cloning
OR
Voice Cloning Ping SignalobjectRequiredBeta

Send ping signal to keep the voice-cloning WebSocket connection alive

Receive

Voice Cloning Audio OutputobjectRequired

Receive audio chunks from the voice-cloning WebSocket

OR
Voice Cloning Event NotificationobjectRequired

Receive completion event notifications from the voice-cloning WebSocket (if send_completion_event is enabled)

OR
Voice Cloning Error ResponseobjectRequired

Receive error messages from the voice-cloning WebSocket