openvino_genai.OmniTalkerSpeechConfig#

class openvino_genai.OmniTalkerSpeechConfig#

Bases: pybind11_object

Standalone speech-side generation config for the Qwen3-Omni talker. Does NOT inherit from GenerationConfig — the thinker text decode is steered by a separate GenerationConfig argument to OmniPipeline.generate. This struct only carries fields the talker actually consumes:

Parameters:

return_audio (bool) – Enable speech output. Default True. Set False to short-circuit the talker and produce text only.
speaker (str | openvino.Tensor) – Speaker identity — either a name (str) looked up in talker_config.speaker_id, or an explicit embedding tensor ([1, 1, talker_hidden_size], f32). Empty string selects the model’s default.
audio_chunk_frames (int) – Number of codec frames accumulated before streaming each audio chunk. Must be >= 1. Each frame is 80ms of audio at 24 kHz (1920 samples).
max_new_tokens (int) – Cap on talker AR steps. Independent of text_config.max_new_tokens (which caps the thinker text decode). The talker pipeline takes the min of this value and the model’s talker_config.talker_max_new_tokens.
rng_seed (int) – RNG seed for deterministic talker + CodePredictor sampling.
talker_repetition_penalty (float | None) – Talker sampling overrides. None = keep the checkpoint default loaded from generation_config.json.
cp_repetition_penalty (float | None) – CodePredictor sampling overrides. Same semantics as talker_*.

__init__(*args, **kwargs)#

Overloaded function.

__init__(self: openvino_genai.py_openvino_genai.OmniTalkerSpeechConfig) -> None
__init__(self: openvino_genai.py_openvino_genai.OmniTalkerSpeechConfig, models_path: os.PathLike | str | bytes) -> None

folder with config.json (talker_config) for default speaker resolution

Methods

`__delattr__`(name, /)	Implement delattr(self, name).
`__dir__`()	Default dir() implementation.
`__eq__`(value, /)	Return self==value.
`__format__`(format_spec, /)	Default object formatter.
`__ge__`(value, /)	Return self>=value.
`__getattribute__`(name, /)	Return getattr(self, name).
`__getstate__`()	Helper for pickle.
`__gt__`(value, /)	Return self>value.
`__hash__`()	Return hash(self).
`__init__`(args, *kwargs)	Overloaded function.
`__init_subclass__`	This method is called when a class is subclassed.
`__le__`(value, /)	Return self<=value.
`__lt__`(value, /)	Return self<value.
`__ne__`(value, /)	Return self!=value.
`__new__`(**kwargs)
`__reduce__`()	Helper for pickle.
`__reduce_ex__`(protocol, /)	Helper for pickle.
`__repr__`()	Return repr(self).
`__setattr__`(name, value, /)	Implement setattr(self, name, value).
`__sizeof__`()	Size of object in memory, in bytes.
`__str__`()	Return str(self).
`__subclasshook__`	Abstract classes can override this to customize issubclass().
`_pybind11_conduit_v1_`

Attributes

`__annotations__`
`audio_chunk_frames`
`cp_repetition_penalty`
`cp_temperature`
`cp_top_k`
`max_new_tokens`
`return_audio`
`rng_seed`
`speaker`	Speaker identity: a name (str) looked up in talker_config.speaker_id, or an explicit embedding tensor ([1, 1, talker_hidden_size], f32).
`speaker_embedding`	Legacy alias.
`talker_repetition_penalty`
`talker_temperature`
`talker_top_k`

__annotations__ = {}#

__class__#: alias of pybind11_type

__delattr__(name, /)#: Implement delattr(self, name).

__dir__()#: Default dir() implementation.

__eq__(value, /)#: Return self==value.

__format__(format_spec, /)#

Default object formatter.

Return str(self) if format_spec is empty. Raise TypeError otherwise.

__ge__(value, /)#: Return self>=value.

__getattribute__(name, /)#: Return getattr(self, name).

__getstate__()#: Helper for pickle.

__gt__(value, /)#: Return self>value.

__hash__()#: Return hash(self).

__init__(*args, **kwargs)#

Overloaded function.

__init__(self: openvino_genai.py_openvino_genai.OmniTalkerSpeechConfig) -> None
__init__(self: openvino_genai.py_openvino_genai.OmniTalkerSpeechConfig, models_path: os.PathLike | str | bytes) -> None

folder with config.json (talker_config) for default speaker resolution

__init_subclass__()#

This method is called when a class is subclassed.

The default implementation does nothing. It may be overridden to extend subclasses.

__le__(value, /)#: Return self<=value.

__lt__(value, /)#: Return self<value.

__ne__(value, /)#: Return self!=value.

__new__(**kwargs)#

__reduce__()#: Helper for pickle.

__reduce_ex__(protocol, /)#: Helper for pickle.

__repr__()#: Return repr(self).

__setattr__(name, value, /)#: Implement setattr(self, name, value).

__sizeof__()#: Size of object in memory, in bytes.

__str__()#: Return str(self).

__subclasshook__()#

Abstract classes can override this to customize issubclass().

This is invoked early on by abc.ABCMeta.__subclasscheck__(). It should return True, False or NotImplemented. If it returns NotImplemented, the normal algorithm is used. Otherwise, it overrides the normal algorithm (and the outcome is cached).

_pybind11_conduit_v1_()#

property audio_chunk_frames#

property cp_repetition_penalty#

property cp_temperature#

property cp_top_k#

property max_new_tokens#

property return_audio#

property rng_seed#

property speaker#: Speaker identity: a name (str) looked up in talker_config.speaker_id, or an explicit embedding tensor ([1, 1, talker_hidden_size], f32).

property speaker_embedding#: Legacy alias. Reading returns the Tensor if speaker holds one, else None. Writing sets the Tensor alternative of the speaker variant.

property talker_repetition_penalty#

property talker_temperature#

property talker_top_k#