类class agentscope.tts

DashScopeRealtimeTTSModel

class DashScopeRealtimeTTSModel(TTSModelBase)

源码Source agentscope/tts/_dashscope/_realtime_model.py:184–468

DashScope 实时 TTS 模型，处理低延迟的实时语音与多模态交互。位于 agentscope.tts 包，继承自 TTSModelBase。核心成员包括 connect、close、push、synthesize。

This model supports streaming input: text can be pushed incrementally via :meth:push, and :meth:synthesize finalizes the current utterance.

For more details see the official document <https://bailian.console.aliyun.com/?tab=doc#/doc/?type=model&url=2938790>_.

嵌套类摘要Nested Class Summary

类Class	类型Kind	说明Description
`Parameters`	类class	参数，处理低延迟的实时语音与多模态交互。 Frontend-exposed parameters for DashScope Realtime TTS models.

构造器摘要Constructor Summary

构造器Constructor	说明Description
`def __init__(credential:` `DashScopeCredential, model: str = 'qwen3-tts-flash-realtime', parameters: 'DashScopeRealtimeTT…`	初始化 `DashScopeRealtimeTTSModel` 实例。 Initialize the DashScope Realtime TTS model.

属性摘要Attribute Summary

属性Attribute	类型Type	说明Description	值Value
`type`	Literal['dashscope_realtime_tts']	类型字段。	`'dashscope_realtime_tts'`
`realtime`	bool	实时字段。	`True`

方法摘要Method Summary

方法Method	说明Description
`async def connect() -> None`async	建立连接并完成初始化。 Establish the WebSocket connection.
`async def close() -> None`async	关闭连接并释放相关资源。 Close the WebSocket connection.
`async def push(text: str, **kwargs: Any) ->` `TTSResponse`async	执行推送相关操作。 Push an incremental text delta for realtime synthesis.
`async def synthesize(text: str \| None = None, **kwargs: Any) ->` `TTSResponse \| AsyncGenerator[TTSResponse, None]`async	执行 `synthesize` 方法。 Finalize synthesis for the current utterance.

构造器详情Constructor Detail

DashScopeRealtimeTTSModel

def __init__(credential: DashScopeCredential, model: str = 'qwen3-tts-flash-realtime', parameters: 'DashScopeRealtimeTTSModel.Parameters | None' = None, stream: bool = True, cold_start_length: int | None = None, cold_start_words: int | None = None, max_retries: int = 3, retry_delay: float = 5.0) -> None

初始化 DashScopeRealtimeTTSModel 实例。

Initialize the DashScope Realtime TTS model.

参数Parameters

名称Name	类型Type	说明Description
`credential`	DashScopeCredential	DashScope 凭证。The DashScope credential.
`model`默认default `'qwen3-tts-flash-realtime'`	str	实时 TTS 模型名称。The realtime TTS model name.
`parameters`默认default `None`	'DashScopeRealtimeTTSModel.Parameters \| None'	参数（类型 `'DashScopeRealtimeTTSModel.Parameters \| None'`）。The TTS parameters (voice, etc.).
`stream`默认default `True`	bool	流（类型 `bool`）。Whether :meth:`synthesize` returns a streaming async generator.
`cold_start_length`默认default `None`	int \| None	cold 启动长度（类型 `int \| None`）。Minimum character count before the first text chunk is sent.
`cold_start_words`默认default `None`	int \| None	cold 启动 words（类型 `int \| None`）。Minimum word count before the first text chunk is sent.
`max_retries`默认default `3`	int	最大重试次数（类型 `int`）。Max retry attempts on WebSocket failure.
`retry_delay`默认default `5.0`	float	重试延迟（类型 `float`）。Initial retry delay in seconds (exponential backoff).

源码Source agentscope/tts/_dashscope/_realtime_model.py:209–256

属性详情Attribute Detail

type

Literal['dashscope_realtime_tts'] = 'dashscope_realtime_tts'

类型字段。声明为 Literal['dashscope_realtime_tts']。

The type field.

源码Source agentscope/tts/_dashscope/_realtime_model.py:205

realtime

bool = True

实时字段。声明为 bool。

The realtime field.

源码Source agentscope/tts/_dashscope/_realtime_model.py:207

方法详情Method Detail

connect

async def connect() -> None

async

建立连接并完成初始化。

Establish the WebSocket connection.

返回Returns None

无返回值。No value is returned.

源码Source agentscope/tts/_dashscope/_realtime_model.py:272–283

close

async def close() -> None

async

关闭连接并释放相关资源。

Close the WebSocket connection.

返回Returns None

无返回值。No value is returned.

源码Source agentscope/tts/_dashscope/_realtime_model.py:285–293

push

async def push(text: str, **kwargs: Any) -> TTSResponse

async

执行推送相关操作。

Push an incremental text delta for realtime synthesis.

参数Parameters

名称Name	类型Type	说明Description
`text`	str	文本（类型 `str`）。An incremental text chunk (delta) to append.
`kwargs`	Any	关键字参数（类型 `Any`）。The kwargs.

返回Returns TTSResponse

返回TTSResponse。`TTSResponse`:

源码Source agentscope/tts/_dashscope/_realtime_model.py:306–363

synthesize

async def synthesize(text: str | None = None, **kwargs: Any) -> TTSResponse | AsyncGenerator[TTSResponse, None]

async

执行 synthesize 方法。

If text was previously pushed via :meth:push, this flushes any remaining buffered text, commits, and waits for audio.

参数Parameters

名称Name	类型Type	说明Description
`text`默认default `None`	str \| None	文本（类型 `str \| None`）。Optional additional text to append before finalizing. If `None`, finalizes previously pushed text.
`kwargs`	Any	关键字参数（类型 `Any`）。The kwargs.

返回Returns TTSResponse | AsyncGenerator[TTSResponse, None]

返回TTSResponse|AsyncGenerator[TTSResponse,None]。`TTSResponse | AsyncGenerator[TTSResponse, None]`:

源码Source agentscope/tts/_dashscope/_realtime_model.py:365–462

嵌套类详情Nested Class Detail

Parameters

class Parameters

参数，处理低延迟的实时语音与多模态交互。位于 agentscope.tts 包。

Frontend-exposed parameters for DashScope Realtime TTS models. Declared in agentscope.tts.

继承成员Inherited Members

继承自Inherited from TTSModelBase

__aenter__, __aexit__, list_models, credential, model, parameters, stream