Configure real-time extraction of the input voice or use a pre-created voice, and consult the pre-created voice API.
qwen3.8-livetranslate-flash-realtime and qwen3.5-livetranslate-flash-realtime can extract the voice from input audio during translation or use a fixed voice created in advance. Real-time voice cloning has no additional cloning charge. Creating a voice in advance is billed separately. Model calls are still billed by token usage. For details, see Billing.
Configure the session
Set enable_voice_clone, voice_clone_options.frequency, and voice in session.update. The following example uses qwen3.8-livetranslate-flash-realtime over WebSocket. For qwen3.5-livetranslate-flash-realtime, replace output_modalities with modalities. The other voice cloning parameters are the same.
| Method | frequency | voice | Description |
|---|---|---|---|
| Clone once in real time | once | default | Extract the voice once from the input audio at the start of the session and reuse it for subsequent output. |
| Clone continuously in real time | always | default | Extract the voice from the input audio before each output. Suitable for multi-party conversations. |
| Use a pre-created voice | never | Pre-created voice ID | Use a fixed voice created in advance without extracting a new voice during this translation. |
When frequency is once or always, voice must be default. The target model of a pre-created voice must match the session model.
One-time real-time cloning
{
"type": "session.update",
"session": {
"output_modalities": ["text", "audio"],
"voice": "default",
"translation": {"language": "en"},
"enable_voice_clone": true,
"voice_clone_options": {"frequency": "once"}
}
}
Continuous real-time cloning
{
"type": "session.update",
"session": {
"output_modalities": ["text", "audio"],
"voice": "default",
"translation": {"language": "en"},
"enable_voice_clone": true,
"voice_clone_options": {"frequency": "always"}
}
}
Use pre-built voice
{
"type": "session.update",
"session": {
"output_modalities": ["text", "audio"],
"voice": "VOICE_ID",
"translation": {"language": "en"},
"enable_voice_clone": true,
"voice_clone_options": {"frequency": "never"}
}
}
Create a pre-built voice
Use Create Pre-built Voice API. target_model must be consistent with the model used in subsequent translation sessions.
Request body
See Pre-built Voice API for field definitions.
Request examples
For a complete request and response example, see Create Pre-built Voice.
Delete a Pre-built Voice
See Delete Pre-built Voice API.
For more interaction steps, see Real-time Speech/Audio-Visual Translation.