PublicMirrors/voice-changer
1
0
Fork 0
mirror of https://github.com/w-okada/voice-changer.git synced 2025-02-02 16:23:58 +03:00
Code Issues Actions Packages Projects Releases Wiki Activity

apply sola for all vc forcely

Browse Source
This commit is contained in:
wataru 2023-04-15 04:58:56 +09:00
parent b3a95b2f7e
commit d85bbae478
16 changed files with 14 additions and 230 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
client/demo/dist/assets/gui_settings/MMVCv13.json vendored
View File

@ -145,10 +145,6 @@
{
"name": "trancateNumThreshold",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

4
client/demo/dist/assets/gui_settings/MMVCv15.json vendored
View File

@ -147,10 +147,6 @@
{
"name": "trancateNumThreshold",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

4
client/demo/dist/assets/gui_settings/RVC.json vendored
View File

@ -166,10 +166,6 @@
{
"name": "rvcQuality",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

4
client/demo/dist/assets/gui_settings/so-vits-svc-40.json vendored
View File

@ -157,10 +157,6 @@
{
"name": "trancateNumThreshold",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

4
client/demo/dist/assets/gui_settings/so-vits-svc-40v2.json vendored
View File

@ -157,10 +157,6 @@
{
"name": "trancateNumThreshold",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

4
client/demo/public/assets/gui_settings/MMVCv13.json
View File

@ -145,10 +145,6 @@
{
"name": "trancateNumThreshold",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

4
client/demo/public/assets/gui_settings/MMVCv15.json
View File

@ -147,10 +147,6 @@
{
"name": "trancateNumThreshold",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

4
client/demo/public/assets/gui_settings/RVC.json
View File

@ -166,10 +166,6 @@
{
"name": "rvcQuality",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

4
client/demo/public/assets/gui_settings/so-vits-svc-40.json
View File

@ -157,10 +157,6 @@
{
"name": "trancateNumThreshold",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

4
client/demo/public/assets/gui_settings/so-vits-svc-40v2.json
View File

@ -157,10 +157,6 @@
{
"name": "trancateNumThreshold",
"options": {}
},
{
"name": "solaEnable",
"options": {}
}
]
},

33
server/voice_changer/MMVCv13/MMVCv13.py
View File

@ -130,7 +130,7 @@ class MMVCv13:
spec = torch.squeeze(spec, 0)
return spec
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaEnabled: bool = False, solaSearchFrame: int = 0):
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaSearchFrame: int = 0):
newData = newData.astype(np.float32) / self.hps.data.max_wav_value
if hasattr(self, "audio_buffer"):
@ -138,10 +138,7 @@ class MMVCv13:
else:
self.audio_buffer = newData
if solaEnabled:
convertSize = inputSize + crossfadeSize + solaSearchFrame
else:
convertSize = inputSize + crossfadeSize
convertSize = inputSize + crossfadeSize + solaSearchFrame
if convertSize < 8192:
convertSize = 8192
@ -160,32 +157,6 @@ class MMVCv13:
return data
def generate_input_old(self, newData: any, inputSize: int, crossfadeSize: int):
newData = newData.astype(np.float32) / self.hps.data.max_wav_value
if hasattr(self, "audio_buffer"):
self.audio_buffer = np.concatenate([self.audio_buffer, newData], 0) # 過去のデータに連結
else:
self.audio_buffer = newData
convertSize = inputSize + crossfadeSize
if convertSize < 8192:
convertSize = 8192
if convertSize % self.hps.data.hop_length != 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.hps.data.hop_length - (convertSize % self.hps.data.hop_length))
self.audio_buffer = self.audio_buffer[-1 * convertSize:] # 変換対象の部分だけ抽出
audio = torch.FloatTensor(self.audio_buffer)
audio_norm = audio.unsqueeze(0) # unsqueeze
spec = self._get_spec(audio_norm)
sid = torch.LongTensor([int(self.settings.srcId)])
data = (self.text_norm, spec, audio_norm, sid)
data = TextAudioSpeakerCollate()([data])
return data
def _onnx_inference(self, data):
if hasattr(self, "onnx_session") == False or self.onnx_session == None:
print("[Voice Changer] No ONNX session.")

7
server/voice_changer/MMVCv15/MMVCv15.py
View File

@ -166,7 +166,7 @@ class MMVCv15:
spec = torch.squeeze(spec, 0)
return spec
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaEnabled: bool = False, solaSearchFrame: int = 0):
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaSearchFrame: int = 0):
newData = newData.astype(np.float32) / self.hps.data.max_wav_value
if hasattr(self, "audio_buffer"):
@ -174,10 +174,7 @@ class MMVCv15:
else:
self.audio_buffer = newData
if solaEnabled:
convertSize = inputSize + crossfadeSize + solaSearchFrame
else:
convertSize = inputSize + crossfadeSize
convertSize = inputSize + crossfadeSize + solaSearchFrame
if convertSize < 8192:
convertSize = 8192

38
server/voice_changer/RVC/RVC.py
View File

@ -165,7 +165,7 @@ class RVC:
def get_processing_sampling_rate(self):
return self.settings.modelSamplingRate
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaEnabled: bool = False, solaSearchFrame: int = 0):
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaSearchFrame: int = 0):
newData = newData.astype(np.float32) / 32768.0
if hasattr(self, "audio_buffer"):
@ -173,10 +173,7 @@ class RVC:
else:
self.audio_buffer = newData
if solaEnabled:
convertSize = inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
else:
convertSize = inputSize + crossfadeSize + self.settings.extraConvertSize
convertSize = inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
if convertSize % 128 != 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 + (128 - (convertSize % 128))
@ -188,30 +185,6 @@ class RVC:
vol = max(rms, self.prevVol * 0.0)
self.prevVol = vol
return (self.audio_buffer, convertSize, vol, solaEnabled)
def generate_input_old(self, newData: any, inputSize: int, crossfadeSize: int):
newData = newData.astype(np.float32) / 32768.0
if hasattr(self, "audio_buffer"):
self.audio_buffer = np.concatenate([self.audio_buffer, newData], 0) # 過去のデータに連結
else:
self.audio_buffer = newData
convertSize = inputSize + crossfadeSize + self.settings.extraConvertSize
# if convertSize % self.hps.data.hop_length != 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。
if convertSize % 128 != 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.hps.data.hop_length - (convertSize % self.hps.data.hop_length))
convertSize = convertSize + (128 - (convertSize % 128))
self.audio_buffer = self.audio_buffer[-1 * convertSize:] # 変換対象の部分だけ抽出
crop = self.audio_buffer[-1 * (inputSize + crossfadeSize):-1 * (crossfadeSize)]
rms = np.sqrt(np.square(crop).mean(axis=0))
vol = max(rms, self.prevVol * 0.0)
self.prevVol = vol
return (self.audio_buffer, convertSize, vol)
def _onnx_inference(self, data):
@ -302,12 +275,7 @@ class RVC:
else:
audio = self._pyTorch_inference(data)
sola_enabled = data[3]
if sola_enabled:
return audio
# return audio[self.settings.extraConvertSize:]
else:
return audio
return audio
def __del__(self):
del self.net_g

7
server/voice_changer/SoVitsSvc40/SoVitsSvc40.py
View File

@ -252,7 +252,7 @@ class SoVitsSvc40:
c = c.unsqueeze(0)
return c, f0, uv
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaEnabled: bool = False, solaSearchFrame: int = 0):
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaSearchFrame: int = 0):
newData = newData.astype(np.float32) / self.hps.data.max_wav_value
if hasattr(self, "audio_buffer"):
@ -260,10 +260,7 @@ class SoVitsSvc40:
else:
self.audio_buffer = newData
if solaEnabled:
convertSize = inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
else:
convertSize = inputSize + crossfadeSize + self.settings.extraConvertSize
convertSize = inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
if convertSize % self.hps.data.hop_length != 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.hps.data.hop_length - (convertSize % self.hps.data.hop_length))

9
server/voice_changer/SoVitsSvc40v2/SoVitsSvc40v2.py
View File

@ -215,7 +215,7 @@ class SoVitsSvc40v2:
c = c.unsqueeze(0)
return c, f0, uv
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaEnabled: bool = False, solaSearchFrame: int = 0):
def generate_input(self, newData: any, inputSize: int, crossfadeSize: int, solaSearchFrame: int = 0):
newData = newData.astype(np.float32) / self.hps.data.max_wav_value
if hasattr(self, "audio_buffer"):
@ -223,10 +223,7 @@ class SoVitsSvc40v2:
else:
self.audio_buffer = newData
if solaEnabled:
convertSize = inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
else:
convertSize = inputSize + crossfadeSize + self.settings.extraConvertSize
convertSize = inputSize + crossfadeSize + solaSearchFrame + self.settings.extraConvertSize
if convertSize % self.hps.data.hop_length != 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.hps.data.hop_length - (convertSize % self.hps.data.hop_length))
@ -274,8 +271,6 @@ class SoVitsSvc40v2:
return result
pass
def _pyTorch_inference(self, data):
if hasattr(self, "net_g") == False or self.net_g == None:
print("[Voice Changer] No pyTorch session.")

110
server/voice_changer/VoiceChanger.py
View File

@ -208,13 +208,9 @@ class VoiceChanger():
# receivedData: tuple of short
def on_request(self, receivedData: AudioInOut) -> tuple[AudioInOut, list[Union[int, float]]]:
if self.settings.solaEnabled:
return self.on_request_sola(receivedData)
else:
return self.on_request_legacy(receivedData)
return self.on_request_sola(receivedData)
def on_request_sola(self, receivedData: AudioInOut) -> tuple[AudioInOut, list[Union[int, float]]]:
print("processing with sola")
processing_sampling_rate = self.voiceChanger.get_processing_sampling_rate()
# 前処理
@ -230,7 +226,7 @@ class VoiceChanger():
crossfade_frame = min(self.settings.crossFadeOverlapSize, block_frame)
self._generate_strength(crossfade_frame)
data = self.voiceChanger.generate_input(newData, block_frame, crossfade_frame, True, sola_search_frame)
data = self.voiceChanger.generate_input(newData, block_frame, crossfade_frame, sola_search_frame)
preprocess_time = t.secs
# 変換処理
@ -295,110 +291,10 @@ class VoiceChanger():
perf = [preprocess_time, mainprocess_time, postprocess_time]
return outputData, perf
def on_request_legacy(self, receivedData: AudioInOut) -> tuple[AudioInOut, list[Union[int, float]]]:
# print("processing with legacy")
processing_sampling_rate = self.voiceChanger.get_processing_sampling_rate()
print_convert_processing(f"------------ Convert processing.... ------------")
# 前処理
with Timer("pre-process") as t:
with Timer("pre-process") as t1:
if self.settings.inputSampleRate != processing_sampling_rate:
newData = cast(AudioInOut, resampy.resample(receivedData, self.settings.inputSampleRate, processing_sampling_rate))
else:
newData = receivedData
# print("t1::::", t1.secs)
inputSize = newData.shape[0]
crossfadeSize = min(self.settings.crossFadeOverlapSize, inputSize)
print_convert_processing(
f" Input data size: {receivedData.shape[0]}/{self.settings.inputSampleRate}hz {inputSize}/{processing_sampling_rate}hz")
print_convert_processing(
f" Crossfade data size: crossfade:{crossfadeSize}, crossfade setting:{self.settings.crossFadeOverlapSize}, input size:{inputSize}")
print_convert_processing(f" Convert data size of {inputSize + crossfadeSize} (+ extra size)")
print_convert_processing(f" will be cropped:{-1 * (inputSize + crossfadeSize)}, {-1 * (crossfadeSize)}")
self._generate_strength(crossfadeSize)
with Timer("pre-process") as t2:
data = self.voiceChanger.generate_input(newData, inputSize, crossfadeSize)
# print("t2::::", t2.secs)
preprocess_time = t.secs
# 変換処理
with Timer("main-process") as t:
try:
# Inference
audio = self.voiceChanger.inference(data)
if hasattr(self, 'np_prev_audio1') == True:
np.set_printoptions(threshold=10000)
prev_overlap_start = -1 * crossfadeSize
prev_overlap = self.np_prev_audio1[prev_overlap_start:]
cur_overlap_start = -1 * (inputSize + crossfadeSize)
cur_overlap_end = -1 * inputSize
cur_overlap = audio[cur_overlap_start:cur_overlap_end]
print_convert_processing(
f" audio:{audio.shape}, prev_overlap:{prev_overlap.shape}, self.np_prev_strength:{self.np_prev_strength.shape}")
powered_prev = prev_overlap * self.np_prev_strength
print_convert_processing(
f" audio:{audio.shape}, cur_overlap:{cur_overlap.shape}, self.np_cur_strength:{self.np_cur_strength.shape}")
print_convert_processing(f" cur_overlap_strt:{cur_overlap_start}, cur_overlap_end{cur_overlap_end}")
powered_cur = cur_overlap * self.np_cur_strength
powered_result = powered_prev + powered_cur
cur = audio[-1 * inputSize:-1 * crossfadeSize]
result = np.concatenate([powered_result, cur], axis=0)
print_convert_processing(
f" overlap:{crossfadeSize}, current:{cur.shape[0]}, result:{result.shape[0]}... result should be same as input")
if cur.shape[0] != result.shape[0]:
print_convert_processing(f" current and result should be same as input")
else:
result = np.zeros(4096).astype(np.int16)
self.np_prev_audio1 = audio
except Exception as e:
print("VC PROCESSING!!!! EXCEPTION!!!", e)
print(traceback.format_exc())
if hasattr(self, "np_prev_audio1"):
del self.np_prev_audio1
return np.zeros(1).astype(np.int16), [0, 0, 0]
mainprocess_time = t.secs
# 後処理
with Timer("post-process") as t:
result = result.astype(np.int16)
if self.settings.inputSampleRate != processing_sampling_rate:
outputData = cast(AudioInOut, resampy.resample(result, processing_sampling_rate, self.settings.inputSampleRate).astype(np.int16))
else:
outputData = result
# outputData = result
print_convert_processing(
f" Output data size of {result.shape[0]}/{processing_sampling_rate}hz {outputData.shape[0]}/{self.settings.inputSampleRate}hz")
if self.settings.recordIO == 1:
self.ioRecorder.writeInput(receivedData)
self.ioRecorder.writeOutput(outputData.tobytes())
# if receivedData.shape[0] != outputData.shape[0]:
# print(f"Padding, in:{receivedData.shape[0]} out:{outputData.shape[0]}")
# outputData = pad_array(outputData, receivedData.shape[0])
# # print_convert_processing(
# # f" Padded!, Output data size of {result.shape[0]}/{processing_sampling_rate}hz {outputData.shape[0]}/{self.settings.inputSampleRate}hz")
postprocess_time = t.secs
print_convert_processing(f" [fin] Input/Output size:{receivedData.shape[0]},{outputData.shape[0]}")
perf = [preprocess_time, mainprocess_time, postprocess_time]
return outputData, perf
def export2onnx(self):
return self.voiceChanger.export2onnx()
##############
PRINT_CONVERT_PROCESSING: bool = False
# PRINT_CONVERT_PROCESSING = True