diff --git a/apps/web/src/pages/voices/components/VideoExtractModal.tsx b/apps/web/src/pages/voices/components/VideoExtractModal.tsx
index 40292bce7..7b3020b48 100644
--- a/apps/web/src/pages/voices/components/VideoExtractModal.tsx
+++ b/apps/web/src/pages/voices/components/VideoExtractModal.tsx
@@ -30,6 +30,7 @@ const VideoExtractModal: React.FC<VideoExtractModalProps> = ({
       title={<span style={{ fontSize: 16, fontWeight: 600 }}>提取视频配音</span>}
       open={open}
       onCancel={() => {
+        if (inputRef.current) inputRef.current.value = ""
         if (isExtracting) return
         onClose()
       }}
@@ -152,7 +153,7 @@ const VideoExtractModal: React.FC<VideoExtractModalProps> = ({
 
           {isExtracting && (
             <p style={{ textAlign: "center", fontSize: 13, color: "#7c3aed", margin: "12px 0 0" }}>
-              {progress === 100 ? "正在提取人声，请稍候..." : "正在上传视频..."}
+              {"正在提取音频，请稍后..."}
             </p>
           )}
         </div>