diff --git a/packages/domain/sentence_timings.py b/packages/domain/sentence_timings.py
index cfeba7ce7..45446d968 100644
--- a/packages/domain/sentence_timings.py
+++ b/packages/domain/sentence_timings.py
@@ -1,6 +1,6 @@
 """共享的句子时间戳计算工具 — 供 Celery TTS 任务和 /lipsync/tts-preview 同步接口复用.
 
-- `_split_script_into_sentences`: 按标点分句（中英文逗号/句号/问号/感叹号/分号/换行）
+- `split_script_into_sentences`: 按标点分句（中英文句号/问号/感叹号/分号/换行，不含逗号，与前端 sentences.ts 保持一致）
 - `_estimate_sentence_timings_by_chars`: 按字数比例估算（静音检测失败时降级）
 - `_probe_audio_duration`: ffprobe 读取音频时长
 - `compute_sentence_timings`: 基于 ffmpeg silencedetect 精确计算每句起止时间
@@ -19,15 +19,16 @@ logger = logging.getLogger(__name__)
 
 
 def split_script_into_sentences(script_text: str) -> list[str]:
-    """按句号/问号/感叹号/分号/逗号/换行分句（与前端 SENTENCE_SPLIT_RE 一致）.
+    """按句号/问号/感叹号/分号/换行分句（与前端 sentences.ts 的 SENTENCE_SPLIT_RE 一致）.
 
-    中文短视频文案习惯用「，」断小句（如"卖花的叫花无缺，卖姜的叫姜子牙"），
-    必须把逗号也纳入分隔符，否则多句文案会被识别成一整句，导致 B-roll 时间戳错位。
+    仅在句末标点（。！？!?；;）和换行处分句，**不再用逗号（,，）切分**。
+    按逗号切分会把连贯句子拆得过碎，导致 B-roll/口播画面按句插入时句数过多、
+    时长过短，效果不符合预期（Issue #1892）。
     """
     text = (script_text or "").strip()
     if not text:
         return []
-    parts = re.split(r"[。！？!?？！;；,，\n\r]+", text)
+    parts = re.split(r"[。！？!?？！;；\n\r]+", text)
     return [p.strip() for p in parts if p.strip()]
 
 
diff --git a/tests/unit/test_sentence_timings.py b/tests/unit/test_sentence_timings.py
index 3551d03ab..73ed4679f 100644
--- a/tests/unit/test_sentence_timings.py
+++ b/tests/unit/test_sentence_timings.py
@@ -46,6 +46,21 @@ class TestSplitScriptIntoSentences(unittest.TestCase):
         result = _split_script_into_sentences("没有标点的句子")
         self.assertEqual(result, ["没有标点的句子"])
 
+    def test_chinese_comma_not_split(self):
+        """Issue #1892：中文逗号「，」不应分句，保持完整小句。"""
+        result = _split_script_into_sentences("卖花的叫花无缺，卖姜的叫姜子牙。")
+        self.assertEqual(result, ["卖花的叫花无缺，卖姜的叫姜子牙"])
+
+    def test_english_comma_not_split(self):
+        """Issue #1892：英文逗号「,」不应分句。"""
+        result = _split_script_into_sentences("Hello, how are you?")
+        self.assertEqual(result, ["Hello, how are you"])
+
+    def test_comma_between_sentences_kept_in_sentence(self):
+        """两个句号间带逗号的长句：逗号不切，只按句号切。"""
+        result = _split_script_into_sentences("第一句，带逗号。第二句，也带逗号！")
+        self.assertEqual(result, ["第一句，带逗号", "第二句，也带逗号"])
+
 
 class TestEstimateSentenceTimingsByChars(unittest.TestCase):
     """Tests for _estimate_sentence_timings_by_chars."""