diff --git a/sidecar/main.py b/sidecar/main.py
index 70051ca..78d9e51 100644
--- a/sidecar/main.py
+++ b/sidecar/main.py
@@ -222,7 +222,16 @@ async def transcribe(
             transcribe_kwargs["initial_prompt"] = initial_prompt
 
         # faster-whisper 전사 실행
-        segments_iter, info = _model.transcribe(audio_array, **transcribe_kwargs)
+        # VAD가 전체 오디오를 제거하면 max() 에러 발생 → VAD 없이 재시도
+        try:
+            segments_iter, info = _model.transcribe(audio_array, **transcribe_kwargs)
+        except ValueError as ve:
+            if "empty sequence" in str(ve) and transcribe_kwargs.get("vad_filter"):
+                logger.warning("VAD가 전체 오디오를 제거함 → VAD 없이 재시도")
+                transcribe_kwargs["vad_filter"] = False
+                segments_iter, info = _model.transcribe(audio_array, **transcribe_kwargs)
+            else:
+                raise
 
         # 세그먼트 수집
         segments_list: list[dict] = []