diff --git a/funasr/auto/auto_model.py b/funasr/auto/auto_model.py
index 6546f7c80..2d3ca92b6 100644
--- a/funasr/auto/auto_model.py
+++ b/funasr/auto/auto_model.py
@@ -1130,6 +1130,21 @@ def inference_with_vad(self, input, input_len=None, **cfg):
             elif kwargs.get("sentence_timestamp", False):
                 if not len(result["text"].strip()):
                     sentence_list = []
+                elif self.punc_model is None and punc_res is None and not sentence_timestamps:
+                    sentence_list = []
+                    for rest, vadsegment in zip(restored_data, vadsegments):
+                        text = str(rest.get("text", "")).strip()
+                        if not text:
+                            continue
+                        sentence_list.append(
+                            {
+                                "start": vadsegment[0],
+                                "end": vadsegment[1],
+                                "text": text,
+                                "sentence": text,
+                                "timestamp": [],
+                            }
+                        )
                 elif punc_res is None:
                     logging.warning(
                         "punc_model is required for sentence_timestamp, skipping sentence segmentation."
diff --git a/tests/test_punc_model_none.py b/tests/test_punc_model_none.py
index 26e354c72..97303b104 100644
--- a/tests/test_punc_model_none.py
+++ b/tests/test_punc_model_none.py
@@ -87,6 +87,66 @@ def test_sentence_timestamp_with_punc_model_none(self, mock_prep, mock_load, moc
         # sentence_info should be empty list since punc_res is unavailable
         self.assertEqual(results[0].get("sentence_info"), [])
 
+    @patch("funasr.auto.auto_model.slice_padding_audio_samples")
+    @patch("funasr.auto.auto_model.load_audio_text_image_video")
+    @patch("funasr.auto.auto_model.prepare_data_iterator")
+    def test_sentence_timestamp_uses_vad_segments_when_asr_has_no_timestamps(
+        self, mock_prep, mock_load, mock_slice
+    ):
+        """SenseVoice should expose its per-VAD text as timestamped sentences."""
+        am = self._make_auto_model(punc_model=None)
+        results_seq = [
+            [{"key": "test_utt", "value": [[120, 800], [1050, 1900]]}],
+            [{"text": "first phrase"}],
+            [{"text": "second phrase"}],
+        ]
+        am.inference = MagicMock(side_effect=lambda *args, **kwargs: results_seq.pop(0))
+        mock_prep.return_value = (["test_utt"], [np.zeros(32000, dtype=np.float32)])
+        mock_load.return_value = np.zeros(32000, dtype=np.float32)
+        mock_slice.return_value = ([np.zeros(16000, dtype=np.float32)], [16000])
+
+        results = am.inference_with_vad("dummy_input", sentence_timestamp=True)
+
+        self.assertEqual(
+            results[0]["sentence_info"],
+            [
+                {
+                    "start": 120,
+                    "end": 800,
+                    "text": "first phrase",
+                    "sentence": "first phrase",
+                    "timestamp": [],
+                },
+                {
+                    "start": 1050,
+                    "end": 1900,
+                    "text": "second phrase",
+                    "sentence": "second phrase",
+                    "timestamp": [],
+                },
+            ],
+        )
+
+    @patch("funasr.auto.auto_model.slice_padding_audio_samples")
+    @patch("funasr.auto.auto_model.load_audio_text_image_video")
+    @patch("funasr.auto.auto_model.prepare_data_iterator")
+    def test_vad_fallback_requires_punc_model_to_be_absent(
+        self, mock_prep, mock_load, mock_slice
+    ):
+        am = self._make_auto_model(punc_model=MagicMock())
+        results_seq = [
+            [{"key": "test_utt", "value": [[120, 800]]}],
+            [{"text": "phrase", "timestamps": []}],
+        ]
+        am.inference = MagicMock(side_effect=lambda *args, **kwargs: results_seq.pop(0))
+        mock_prep.return_value = (["test_utt"], [np.zeros(16000, dtype=np.float32)])
+        mock_load.return_value = np.zeros(16000, dtype=np.float32)
+        mock_slice.return_value = ([np.zeros(16000, dtype=np.float32)], [16000])
+
+        results = am.inference_with_vad("dummy_input", sentence_timestamp=True)
+
+        self.assertEqual(results[0]["sentence_info"], [])
+
     @patch("funasr.auto.auto_model.slice_padding_audio_samples")
     @patch("funasr.auto.auto_model.load_audio_text_image_video")
     @patch("funasr.auto.auto_model.prepare_data_iterator")