Update pipeline/postprocess.py

pipeline/postprocess.py

@@ -3,7 +3,7 @@
 MIN_DISPLAY     = 2.5
 MAX_DISPLAY     = 7.0
 MERGE_MAX_CHARS = 72
-SPLIT_MAX_CHARS = 60
+SPLIT_MAX_CHARS = 80
 MAX_GAP         = 1.5
 
 
@@ -108,11 +108,12 @@ def apply_min_display(segments: list) -> list:
     return result
 
 
-def split_long_segment(seg: dict) -> list:
+def split_long_segment(seg: dict, depth: int = 0) -> list:
     words = seg.get("words", [])
     text  = seg["text"].strip()
 
-    if len(text) <= SPLIT_MAX_CHARS or len(words) < 3:
+    # حداکثر ۲ بار split — جلوگیری از تک کلمه شدن
+    if depth >= 2 or len(text) <= SPLIT_MAX_CHARS or len(words) < 4:
         return [seg]
 
     best_idx = len(words) // 2
@@ -145,7 +146,7 @@ def split_long_segment(seg: dict) -> list:
         "words": words2,
     }
 
-    return split_long_segment(seg1) + split_long_segment(seg2)
+    return split_long_segment(seg1, depth + 1) + split_long_segment(seg2, depth + 1)
 
 
 def build_srt(segments: list) -> str: