From f8823b1913c29ce2710f92d51b74cb84b74323b0 Mon Sep 17 00:00:00 2001 From: Hiroshiba Kazuyuki Date: Sun, 11 Mar 2018 02:37:25 +0900 Subject: 無音を無視するように変更 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- become_yukarin/config/config.py | 3 +++ become_yukarin/voice_changer.py | 9 +++++---- 2 files changed, 8 insertions(+), 4 deletions(-) (limited to 'become_yukarin') diff --git a/become_yukarin/config/config.py b/become_yukarin/config/config.py index f1f24cf..ed128c4 100644 --- a/become_yukarin/config/config.py +++ b/become_yukarin/config/config.py @@ -145,3 +145,6 @@ def backward_compatible(d: Dict): d['model']['generator_extensive_layers'] = 8 d['model']['discriminator_base_channels'] = 32 d['model']['discriminator_extensive_layers'] = 5 + + if 'weak_discriminator' not in d['model']: + d['model']['weak_discriminator'] = False diff --git a/become_yukarin/voice_changer.py b/become_yukarin/voice_changer.py index bed155f..7f7bbe4 100644 --- a/become_yukarin/voice_changer.py +++ b/become_yukarin/voice_changer.py @@ -4,6 +4,7 @@ from typing import NamedTuple import numpy +from become_yukarin.param import Param from .acoustic_converter import AcousticConverter from .data_struct import AcousticFeature from .data_struct import Wave @@ -78,10 +79,12 @@ class VoiceChangerStream(object): self, sampling_rate: int, frame_period: float, + order: int, in_dtype=numpy.float32, ): self.sampling_rate = sampling_rate self.frame_period = frame_period + self.order = order self.in_dtype = in_dtype self.voice_changer: VoiceChanger = None @@ -189,8 +192,7 @@ class VoiceChangerStream(object): return in_feature def convert(self, start_time: float, time_length: float, extra_time: float): - order = self.voice_changer.acoustic_converter.config.dataset.param.acoustic_feature_param.order - sizes = AcousticFeature.get_sizes(sampling_rate=self.sampling_rate, order=order) + sizes = AcousticFeature.get_sizes(sampling_rate=self.sampling_rate, order=self.order) keys = ['f0', 'aperiodicity', 'mfcc', 'voiced'] in_feature = self.fetch( start_time=start_time, @@ -209,8 +211,7 @@ class VoiceChangerStream(object): return out_feature def post_convert(self, start_time: float, time_length: float): - order = self.voice_changer.acoustic_converter.config.dataset.param.acoustic_feature_param.order - sizes = AcousticFeature.get_sizes(sampling_rate=self.sampling_rate, order=order) + sizes = AcousticFeature.get_sizes(sampling_rate=self.sampling_rate, order=self.order) keys = ['f0', 'aperiodicity', 'spectrogram', 'voiced'] out_feature = self.fetch( start_time=start_time, -- cgit v1.2.3-70-g09d2