1berry/wokada-voice-changer
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

5. mv volume

Browse Source
This commit is contained in:
wataru 2023-03-24 09:47:14 +09:00
parent 0de24ca000
commit 1143ad88fd
1 changed files with 15 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
server/voice_changer/DDSP_SVC/DDSP_SVC.py
View File

@ -72,11 +72,11 @@ class DDSP_SVC:
self.settings.configFile = config
# model
model, args = vo.load_model(pyTorch_model_file)
# hubert
self.model = model
self.args = args
self.hop_size = int(self.args.data.block_size * 44100 / self.args.data.sampling_rate)
# hubert
vec_path = self.params["hubert"]
self.encoder = vo.Units_Encoder(
args.data.encoder,
@ -92,6 +92,8 @@ class DDSP_SVC:
float(50),
float(1100))
self.volume_extractor = vo.Volume_Extractor(self.hop_size)
return self.get_info()
def update_setteings(self, key: str, val: any):
@ -157,7 +159,6 @@ class DDSP_SVC:
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int):
newData = newData.astype(np.float32) / 32768.0
# newData = newData.astype(np.float32) / self.hps.data.max_wav_value
hop_size = int(self.args.data.block_size * 44100 / self.args.data.sampling_rate)
if hasattr(self, "audio_buffer"):
self.audio_buffer = np.concatenate([self.audio_buffer, newData], 0) # 過去のデータに連結
@ -165,9 +166,9 @@ class DDSP_SVC:
self.audio_buffer = newData
convertSize = inputSize + crossfadeSize + self.settings.extraConvertSize
print("hopsize", hop_size)
if convertSize % hop_size != 0: # モデルの出力のホップサイズで切り捨てが発生するので補う。
convertSize = convertSize + (hop_size - (convertSize % hop_size))
print("hopsize", self.hop_size)
if convertSize % self.hop_size != 0: # モデルの出力のホップサイズで切り捨てが発生するので補う。
convertSize = convertSize + (self.hop_size - (convertSize % self.hop_size))
print("convsize", convertSize)
self.audio_buffer = self.audio_buffer[-1 * convertSize:] # 変換対象の部分だけ抽出
@ -176,8 +177,10 @@ class DDSP_SVC:
f0 = torch.from_numpy(f0).float().unsqueeze(-1).unsqueeze(0)
f0 = f0 * 2 ** (float(10) / 12)
volume = self.volume_extractor.extract(self.audio_buffer)
audio = torch.from_numpy(self.audio_buffer).float().unsqueeze(0)
seg_units = self.encoder.encode(audio, 44100, hop_size)
seg_units = self.encoder.encode(audio, 44100, self.hop_size)
print("audio1", audio)
# crop = self.audio_buffer[-1 * (inputSize + crossfadeSize):-1 * (crossfadeSize)]
@ -188,7 +191,7 @@ class DDSP_SVC:
# c, f0 = self.get_unit_f0(self.audio_buffer, self.settings.tran)
# return (c, f0, convertSize, vol)
wavfile.write("tmp2.wav", 44100, (self.audio_buffer * 32768.0).astype(np.int16))
return (seg_units, f0)
return (seg_units, f0, volume)
def _onnx_inference(self, data):
if hasattr(self, "onnx_session") == False or self.onnx_session == None:
@ -259,8 +262,9 @@ class DDSP_SVC:
audio, sample_rate = librosa.load("tmp2.wav", sr=None)
print("SR:", sample_rate)
seg_units = data[0]
c = data[0]
f0 = data[1]
volume = data[2]
if len(audio.shape) > 1:
audio = librosa.to_mono(audio)
@ -268,8 +272,6 @@ class DDSP_SVC:
print("hop_size", hop_size)
volume_extractor = vo.Volume_Extractor(hop_size)
volume = volume_extractor.extract(audio)
mask = (volume > 10 ** (float(-60) / 20)).astype('float')
mask = np.pad(mask, (4, 4), constant_values=(mask[0], mask[-1]))
mask = np.array([np.max(mask[n: n + 9]) for n in range(len(mask) - 8)])
@ -283,12 +285,10 @@ class DDSP_SVC:
with torch.no_grad():
start_frame = 0
seg_f0 = f0
seg_volume = volume
seg_output, _, (s_h, s_n) = self.model(seg_units, seg_f0, seg_volume, spk_id=spk_id, spk_mix_dict=None)
seg_output *= mask[:, start_frame * self.args.data.block_size: (start_frame + seg_units.size(1)) * self.args.data.block_size]
seg_output, _, (s_h, s_n) = self.model(c, f0, seg_volume, spk_id=spk_id, spk_mix_dict=None)
seg_output *= mask[:, start_frame * self.args.data.block_size: (start_frame + c.size(1)) * self.args.data.block_size]
output_sample_rate = self.args.data.sampling_rate