diff --git a/server/voice_changer/DiffusionSVC/DiffusionSVC.py b/server/voice_changer/DiffusionSVC/DiffusionSVC.py
index 7f371eb2..d202c000 100644
--- a/server/voice_changer/DiffusionSVC/DiffusionSVC.py
+++ b/server/voice_changer/DiffusionSVC/DiffusionSVC.py
@@ -104,7 +104,7 @@ class DiffusionSVC(VoiceChangerModel):
             convertSize = convertSize + (128 - (convertSize % 128))
 
         # バッファがたまっていない場合はzeroで補う
-        generateFeatureLength = int(((convertSize / self.inputSampleRate) * self.slotInfo.samplingRate) / 512) + 1        
+        generateFeatureLength = int(((convertSize / self.inputSampleRate) * self.slotInfo.samplingRate) / 512) + 1
         if self.audio_buffer.shape[0] < convertSize:
             self.audio_buffer = np.concatenate([np.zeros([convertSize]), self.audio_buffer])
             self.pitchf_buffer = np.concatenate([np.zeros(generateFeatureLength), self.pitchf_buffer])
@@ -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....")
diff --git a/server/voice_changer/DiffusionSVC/inferencer/DiffusionSVCInferencer.py b/server/voice_changer/DiffusionSVC/inferencer/DiffusionSVCInferencer.py
index eea196c4..f54f024c 100644
--- a/server/voice_changer/DiffusionSVC/inferencer/DiffusionSVCInferencer.py
+++ b/server/voice_changer/DiffusionSVC/inferencer/DiffusionSVCInferencer.py
@@ -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)
 
diff --git a/server/voice_changer/Local/ServerDevice.py b/server/voice_changer/Local/ServerDevice.py
index 2d486956..c6a8dda1 100644
--- a/server/voice_changer/Local/ServerDevice.py
+++ b/server/voice_changer/Local/ServerDevice.py
@@ -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
diff --git a/server/voice_changer/VoiceChangerV2.py b/server/voice_changer/VoiceChangerV2.py
index 1f3c7fe2..5ed0a418 100644
--- a/server/voice_changer/VoiceChangerV2.py
+++ b/server/voice_changer/VoiceChangerV2.py
@@ -260,7 +260,7 @@ class VoiceChangerV2(VoiceChangerIF):
 
                 print_convert_processing(f" Output data size of {result.shape[0]}/{processing_sampling_rate}hz {result .shape[0]}/{self.settings.outputSampleRate}hz")
 
-                if receivedData.shape[0] != result .shape[0]:
+                if receivedData.shape[0] != result.shape[0]:
                     outputData = pad_array(result, receivedData.shape[0])
                     pass
                 else: