update vad parameter

transcribe/pipelines/pipe_vad.py

@@ -21,7 +21,7 @@ class VadPipe(BasePipe):
             fusion_threshold=0.45,        # 提高以更好地融合语音片段
             min_speech_duration=0.2,      # 略微降低以捕获短音节
             max_speech_duration=20,       # 保持不变
-            min_silence_duration=300,     # 增加到300毫秒，允许说话间的自然停顿
+            # min_silence_duration=300,     # 增加到300毫秒，允许说话间的自然停顿
             sample_rate=cls.sample_rate # 采样率，音频信号的采样频率
         )
             cls.vac = FixedVADIterator(cls.model.silero_vad, sampling_rate=cls.sample_rate,)

transcribe/strategy.py

@@ -169,8 +169,15 @@ class TranscriptBuffer:
     def get_seg_id(self) -> int:
         return self._current_seg_id
     
-    def next_seg_id(self) -> int:
-        return self._current_seg_id + 1
+    @property
+    def current_sentences_length(self) -> int:
+        count = 0
+        for item in self._sentences:
+            if self._separator:
+                count += len(item.split(self._separator))
+            else:
+                count += len(item)
+        return count
     
     def update_pending_text(self, text: str) -> None:
         """更新临时缓冲字符串"""
@@ -192,10 +199,13 @@ class TranscriptBuffer:
         
         count = 0
         current_sentences = []
-        while len(self._sentences) and count <=20:
+        while len(self._sentences) and count < 20:
             item = self._sentences.popleft()
             current_sentences.append(item)
-            count += len(item.split(self._separator))
+            if self._separator:
+                count += len(item.split(self._separator))
+            else:
+                count += len(item)
         if current_sentences:
             self._segments.append("".join(current_sentences))
         logger.debug(f"=== count to paragraph ===")