debug cut index

transcribe/strategy.py

@@ -48,7 +48,7 @@ class TranscriptToken:
         return PAUSEE_END_PATTERN.search(self.text) is not  None
 
     def buffer_index(self) -> int:
-        return int(self.t1 / 100 * SAMPLE_RATE)
+        return max(int(self.t1 / 100 * SAMPLE_RATE) - 300, 0)
 
 @dataclass
 class TranscriptChunk:
@@ -108,6 +108,9 @@ class TranscriptChunk:
         return any(seg.is_punctuation() for seg in self.items)
     
     def get_buffer_index(self) -> int:
+        logger.debug("==== Current cut item ====")
+        logger.debug(f"{self.items[-1]}")
+        logger.debug("==========================")
         return self.items[-1].buffer_index()
     
     def is_end_sentence(self) ->bool:
@@ -182,10 +185,13 @@ class TranscriptBuffer:
         self.update_pending_text(stable_string)
         if is_end_sentence:
             self.commit_paragraph(end_of_sentence=True)
+            self.update_pending_text(remaining_string)
             # if len() >=20
+            return True
         else:
             self.commit_line()
-        self.update_pending_text(remaining_string)
+            self.update_pending_text(remaining_string)
+            return False
 
        
     @property
@@ -278,11 +284,10 @@ class TranscriptStabilityAnalyzer:
         logger.debug(f"Current cut index: {frame_cut_index}, Stable string: {stable_str}, Remaining_str:{remaining_str}")
 
         prev_seg_id = self._transcript_buffer.get_seg_id()
-        self._transcript_buffer.update_and_commit(stable_str, remaining_str, is_end_sentence)
-
+        commit_paragraph = self._transcript_buffer.update_and_commit(stable_str, remaining_str, is_end_sentence)
         logger.debug(f"current buffer: {self._transcript_buffer.__dict__}")
 
-        if self._transcript_buffer.get_seg_id() > prev_seg_id:
+        if commit_paragraph:
             # 表示生成了一个新段落 换行
             yield TranscriptResult(
                 seg_id=prev_seg_id,