Skip to content

利用 AI 大模型和自动化工作流,根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.

fix(subtitle): drop transcription cues beyond script

Rudy Celekliba4c6af8ef0c
diff --git a/app/services/subtitle.py b/app/services/subtitle.py
index b94746e7..ef62694e 100644
--- a/app/services/subtitle.py
+++ b/app/services/subtitle.py
@@ -328,6 +328,13 @@ def correct(subtitle_file, video_script):
script_index += 1
corrected = True
+ if subtitle_index < len(subtitle_items):
+ logger.warning(
+ f"Dropping {len(subtitle_items) - subtitle_index} transcription cue(s) "
+ "after the script ends"
+ )
+ corrected = True
+
if corrected:
with open(subtitle_file, "w", encoding="utf-8") as fd:
for i, item in enumerate(new_subtitle_items):
diff --git a/test/services/test_subtitle.py b/test/services/test_subtitle.py
index 452c073d..1fdd129a 100644
--- a/test/services/test_subtitle.py
+++ b/test/services/test_subtitle.py
@@ -259,6 +259,22 @@ class TestSubtitleService(unittest.TestCase):
self.assertEqual(items[0][1], "00:00:00,100 --> 00:00:02,000")
self.assertEqual(items[0][2], "Hello world")
+ def test_correct_removes_transcription_after_script_ends(self):
+ """Whisper's trailing hallucinations must not appear as final captions."""
+ original_srt = (
+ "1\n00:00:00,100 --> 00:00:01,000\nHello world\n\n"
+ "2\n00:00:01,000 --> 00:00:02,000\nUnspoken words\n\n"
+ )
+
+ with tempfile.TemporaryDirectory() as tmp_dir:
+ subtitle_file = Path(tmp_dir) / "subtitle.srt"
+ subtitle_file.write_text(original_srt, encoding="utf-8")
+
+ subtitle.correct(str(subtitle_file), "Hello world")
+ items = subtitle.file_to_subtitles(str(subtitle_file))
+
+ self.assertEqual([item[2] for item in items], ["Hello world"])
+
def test_correct_replaces_mismatch_and_appends_missing_script_line(self):
"""
转写结果与脚本完全不一致时仍应以脚本为准;脚本多出的句子没有可复用