Update prepare_dataset.py

prepare_dataset.py

@@ -40,10 +40,10 @@ def merge_subs(subs: list[Subtitle], max_sample_duration_s: float) -> list[Subti
     previous_sub: Subtitle = subs[0]
     for sub in subs[1:]:
         # there is a short silence between subs or we reach max duration
-        if previous_sub.end < sub.start or previous_sub.duration + sub.duration > max_sample_duration_s:
+        if previous_sub.end + 0.2 < sub.start or previous_sub.duration + sub.duration > max_sample_duration_s:
             merged_subs.append(previous_sub)
             previous_sub = sub
-        elif previous_sub and previous_sub.end == sub.start:
+        elif previous_sub and previous_sub.end + 0.2 >= sub.start:
             previous_sub = Subtitle.merge(previous_sub, sub)
         else:
             raise ValueError("Subtitles are not in order")
@@ -90,5 +90,3 @@ with (dataset_path.parent / "metadata.csv").open('w', encoding="utf-8") as f:
     writer = csv.DictWriter(f, fieldnames=columns)
     writer.writeheader()
     writer.writerows(csv_lines)
-
-