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

bugfix:

Browse Source 
diffusion svc server mode silence front value
This commit is contained in:
w-okada 2023-07-18 08:29:17 +09:00
parent b559582dc4
commit 0957fbc923
4 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -151,7 +151,7 @@ class DiffusionSVC(VoiceChangerModel):
speedUp = self.settings.speedUp
embOutputLayer = 12
useFinalProj = False
silenceFrontSec = self.settings.extraConvertSize / self.slotInfo.samplingRate if self.settings.silenceFront else 0. # extaraConvertSize(既にモデルのサンプリングレートにリサンプリング済み)の秒数。モデルのサンプリングレートで処理(★１)。
silenceFrontSec = self.settings.extraConvertSize / self.inputSampleRate if self.settings.silenceFront else 0. # extaraConvertSize(既にモデルのサンプリングレートにリサンプリング済み)の秒数。モデルのサンプリングレートで処理(★１)。
try:
audio_out, self.pitchf_buffer, self.feature_buffer = self.pipeline.exec(
@ -169,7 +169,6 @@ class DiffusionSVC(VoiceChangerModel):
protect
)
result = audio_out.detach().cpu().numpy()
return result
except DeviceCannotSupportHalfPrecisionException as e: # NOQA
print("[Device Manager] Device cannot support half precision. Fallback to float....")

1
server/voice_changer/DiffusionSVC/inferencer/DiffusionSVCInferencer.py
View File

@ -125,7 +125,6 @@ class DiffusionSVCInferencer(Inferencer):
with Timer("pre-process") as t: # NOQA
start_frame = int(silence_front * self.vocoder.vocoder_sample_rate / self.vocoder.vocoder_hop_size)
out_wav = self.mel2wav(out_mel, pitch, start_frame=start_frame)
out_wav *= mask
# print("[ ----Timer::3: ]", t.secs, start_frame, out_mel.shape)

4
server/voice_changer/Local/ServerDevice.py
View File

@ -271,7 +271,9 @@ class ServerDevice:
# Blockサイズを計算
currentInputChunkNum = self.settings.serverReadChunkSize
block_frame = currentInputChunkNum * 128
# block_frame = currentInputChunkNum * 128
block_frame = int(currentInputChunkNum * 128 * (self.settings.serverInputAudioSampleRate / 48000))
sd.default.blocksize = block_frame
# main loop