PublicMirrors/voice-changer
1
0
Fork 0
mirror of https://github.com/w-okada/voice-changer.git synced 2025-03-12 10:20:11 +03:00
Code Issues Actions Packages Projects Releases Wiki Activity

WIP: support so-vits-svc, message impruvement

Browse Source
This commit is contained in:
wataru 2023-03-11 03:59:03 +09:00
parent 010eb395ef
commit d7cff0709f
2 changed files with 29 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
server/voice_changer/SoVitsSvc40v2/SoVitsSvc40v2.py
View File

@ -141,7 +141,8 @@ class SoVitsSvc40v2:
wav_44k = audio_buffer wav_44k = audio_buffer
# f0 = utils.compute_f0_parselmouth(wav, sampling_rate=self.target_sample, hop_length=self.hop_size) # f0 = utils.compute_f0_parselmouth(wav, sampling_rate=self.target_sample, hop_length=self.hop_size)
f0 = utils.compute_f0_dio(wav_44k, sampling_rate=self.hps.data.sampling_rate, hop_length=self.hps.data.hop_length) f0 = utils.compute_f0_dio(wav_44k, sampling_rate=self.hps.data.sampling_rate, hop_length=self.hps.data.hop_length)
print(f"--- >>>>> ---- >>>> {wav_44k.shape[0] / self.hps.data.hop_length}") if wav_44k.shape[0] % self.hps.data.hop_length != 0:
print(f" !!! !!! !!! wav size not multiple of hopsize: {wav_44k.shape[0] / self.hps.data.hop_length}")
f0, uv = utils.interpolate_f0(f0) f0, uv = utils.interpolate_f0(f0)
f0 = torch.FloatTensor(f0) f0 = torch.FloatTensor(f0)

41
server/voice_changer/VoiceChanger.py
View File

@ -147,7 +147,7 @@ class VoiceChanger():
# receivedData: tuple of short # receivedData: tuple of short
def on_request(self, receivedData: any): def on_request(self, receivedData: any):
processing_sampling_rate = self.voiceChanger.get_processing_sampling_rate() processing_sampling_rate = self.voiceChanger.get_processing_sampling_rate()
print(f"------------ Convert processing.... ------------") print_convert_processing(f"------------ Convert processing.... ------------")
# 前処理 # 前処理
with Timer("pre-process") as t: with Timer("pre-process") as t:
@ -158,7 +158,8 @@ class VoiceChanger():
inputSize = newData.shape[0] inputSize = newData.shape[0]
convertSize = inputSize + min(self.settings.crossFadeOverlapSize, inputSize) convertSize = inputSize + min(self.settings.crossFadeOverlapSize, inputSize)
print(f" Input data size of {receivedData.shape[0]}/{self.settings.inputSampleRate}hz {inputSize}/{processing_sampling_rate}hz") print_convert_processing(
f" Input data size of {receivedData.shape[0]}/{self.settings.inputSampleRate}hz {inputSize}/{processing_sampling_rate}hz")
if convertSize < 8192: if convertSize < 8192:
convertSize = 8192 convertSize = 8192
@ -170,8 +171,9 @@ class VoiceChanger():
overlapSize = min(self.settings.crossFadeOverlapSize, inputSize) overlapSize = min(self.settings.crossFadeOverlapSize, inputSize)
cropRange = (-1 * (inputSize + overlapSize), -1 * overlapSize) cropRange = (-1 * (inputSize + overlapSize), -1 * overlapSize)
print(f" Convert input data size of {convertSize}") print_convert_processing(f" Convert input data size of {convertSize}")
print(f" overlap:{overlapSize}, cropRange:{cropRange}") print_convert_processing(f" overlap:{overlapSize}, cropRange:{cropRange}")
self._generate_strength(inputSize) self._generate_strength(inputSize)
data = self.voiceChanger.generate_input(newData, convertSize, cropRange) data = self.voiceChanger.generate_input(newData, convertSize, cropRange)
preprocess_time = t.secs preprocess_time = t.secs
@ -190,16 +192,18 @@ class VoiceChanger():
cur_overlap = audio[cur_overlap_start:cur_overlap_end] cur_overlap = audio[cur_overlap_start:cur_overlap_end]
# cur_overlap = audio[-1 * (inputSize + overlapSize):-1 * inputSize] # cur_overlap = audio[-1 * (inputSize + overlapSize):-1 * inputSize]
powered_prev = prev_overlap * self.np_prev_strength powered_prev = prev_overlap * self.np_prev_strength
print(f" ---- ---- ---- audio:{audio.shape}, cur_overlap:{cur_overlap.shape}, self.np_cur_strength:{self.np_cur_strength.shape}") print_convert_processing(
print(f" ---- ---- ---------------- {cur_overlap_start}, {cur_overlap_end}") 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_cur = cur_overlap * self.np_cur_strength
powered_result = powered_prev + powered_cur powered_result = powered_prev + powered_cur
cur = audio[-1 * inputSize:-1 * overlapSize] cur = audio[-1 * inputSize:-1 * overlapSize]
result = np.concatenate([powered_result, cur], axis=0) result = np.concatenate([powered_result, cur], axis=0)
print(f" overlap:{overlapSize}, current:{cur.shape[0]}, result:{result.shape[0]}... result should be same as input") print_convert_processing(
# print(prev_overlap.shape, self.np_prev_strength.shape, cur_overlap.shape, self.np_cur_strength.shape) f" overlap:{overlapSize}, current:{cur.shape[0]}, result:{result.shape[0]}... result should be same as input")
# print(">>>>>>>>>>>", -1 * (inputSize + overlapSize), -1 * inputSize, self.np_prev_audio1.shape, overlapSize) if cur.shape[0] != result.shape[0]:
print_convert_processing(f" current and result should be same as input")
else: else:
result = np.zeros(4096).astype(np.int16) result = np.zeros(4096).astype(np.int16)
@ -221,7 +225,8 @@ class VoiceChanger():
else: else:
outputData = result outputData = result
print(f" Output data size of {result.shape[0]}/{processing_sampling_rate}hz {outputData.shape[0]}/{self.settings.inputSampleRate}hz") 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: if self.settings.recordIO == 1:
self.ioRecorder.writeInput(receivedData) self.ioRecorder.writeInput(receivedData)
@ -229,16 +234,26 @@ class VoiceChanger():
if receivedData.shape[0] != outputData.shape[0]: if receivedData.shape[0] != outputData.shape[0]:
outputData = pad_array(outputData, receivedData.shape[0]) outputData = pad_array(outputData, receivedData.shape[0])
print( print_convert_processing(
f" Padded!, Output data size of {result.shape[0]}/{processing_sampling_rate}hz {outputData.shape[0]}/{self.settings.inputSampleRate}hz") f" Padded!, Output data size of {result.shape[0]}/{processing_sampling_rate}hz {outputData.shape[0]}/{self.settings.inputSampleRate}hz")
postprocess_time = t.secs postprocess_time = t.secs
print(" [fin] Input/Output size:", receivedData.shape[0], outputData.shape[0]) print_convert_processing(f" [fin] Input/Output size:{receivedData.shape[0]},{outputData.shape[0]}")
perf = [preprocess_time, mainprocess_time, postprocess_time] perf = [preprocess_time, mainprocess_time, postprocess_time]
return outputData, perf return outputData, perf
##############
PRINT_CONVERT_PROCESSING = False
# PRINT_CONVERT_PROCESSING = True
def print_convert_processing(mess: str):
if PRINT_CONVERT_PROCESSING == True:
print(mess)
def pad_array(arr, target_length): def pad_array(arr, target_length):
current_length = arr.shape[0] current_length = arr.shape[0]
if current_length >= target_length: if current_length >= target_length:
@ -250,8 +265,6 @@ def pad_array(arr, target_length):
padded_arr = np.pad(arr, (pad_left, pad_right), 'constant', constant_values=(0, 0)) padded_arr = np.pad(arr, (pad_left, pad_right), 'constant', constant_values=(0, 0))
return padded_arr return padded_arr
##############
class Timer(object): class Timer(object):
def __init__(self, title: str): def __init__(self, title: str):