|
|
@ -907,7 +907,9 @@ class StyleFastSpeech2Inference(FastSpeech2Inference):
|
|
|
|
energy: Union[paddle.Tensor, np.ndarray]=None,
|
|
|
|
energy: Union[paddle.Tensor, np.ndarray]=None,
|
|
|
|
energy_scale: Union[int, float]=None,
|
|
|
|
energy_scale: Union[int, float]=None,
|
|
|
|
energy_bias: Union[int, float]=None,
|
|
|
|
energy_bias: Union[int, float]=None,
|
|
|
|
robot: bool=False):
|
|
|
|
robot: bool=False,
|
|
|
|
|
|
|
|
spk_emb=None,
|
|
|
|
|
|
|
|
spk_id=None):
|
|
|
|
"""
|
|
|
|
"""
|
|
|
|
Parameters
|
|
|
|
Parameters
|
|
|
|
----------
|
|
|
|
----------
|
|
|
@ -938,8 +940,9 @@ class StyleFastSpeech2Inference(FastSpeech2Inference):
|
|
|
|
Tensor
|
|
|
|
Tensor
|
|
|
|
Output sequence of features (L, odim).
|
|
|
|
Output sequence of features (L, odim).
|
|
|
|
"""
|
|
|
|
"""
|
|
|
|
|
|
|
|
spk_id = paddle.to_tensor(spk_id)
|
|
|
|
normalized_mel, d_outs, p_outs, e_outs = self.acoustic_model.inference(
|
|
|
|
normalized_mel, d_outs, p_outs, e_outs = self.acoustic_model.inference(
|
|
|
|
text, durations=None, pitch=None, energy=None)
|
|
|
|
text, durations=None, pitch=None, energy=None, spk_emb=spk_emb, spk_id=spk_id)
|
|
|
|
# priority: groundtruth > scale/bias > previous output
|
|
|
|
# priority: groundtruth > scale/bias > previous output
|
|
|
|
# set durations
|
|
|
|
# set durations
|
|
|
|
if isinstance(durations, np.ndarray):
|
|
|
|
if isinstance(durations, np.ndarray):
|
|
|
@ -991,7 +994,10 @@ class StyleFastSpeech2Inference(FastSpeech2Inference):
|
|
|
|
durations=durations,
|
|
|
|
durations=durations,
|
|
|
|
pitch=pitch,
|
|
|
|
pitch=pitch,
|
|
|
|
energy=energy,
|
|
|
|
energy=energy,
|
|
|
|
use_teacher_forcing=True)
|
|
|
|
use_teacher_forcing=True,
|
|
|
|
|
|
|
|
spk_emb=spk_emb,
|
|
|
|
|
|
|
|
spk_id=spk_id
|
|
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
logmel = self.normalizer.inverse(normalized_mel)
|
|
|
|
logmel = self.normalizer.inverse(normalized_mel)
|
|
|
|
return logmel
|
|
|
|
return logmel
|
|
|
|