diff --git a/components/SmartChatBox/InputBar.tsx b/components/SmartChatBox/InputBar.tsx
index 3e8fc12..1785d0a 100644
--- a/components/SmartChatBox/InputBar.tsx
+++ b/components/SmartChatBox/InputBar.tsx
@@ -23,9 +23,10 @@ interface InputBarProps {
   setVideoPreview?: (url: string, id: string) => void;
   initialVideoUrl?: string;
   initialVideoId?: string;
+  setIsFocusChatInput?: (v: boolean) => void;
 }
 
-export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVideoId }: InputBarProps) {
+export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVideoId, setIsFocusChatInput }: InputBarProps) {
   const [text, setText] = useState("");
   const [isUploading, setIsUploading] = useState(false);
   const [uploadProgress, setUploadProgress] = useState(0);
@@ -285,6 +286,8 @@ export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVide
                     handleSend();
                   }
                 }}
+                onFocus={() => setIsFocusChatInput?.(true)}
+                onBlur={() => setIsFocusChatInput?.(false)}
                 data-alt="text-input"
               />
             </motion.div>
diff --git a/components/SmartChatBox/SmartChatBox.tsx b/components/SmartChatBox/SmartChatBox.tsx
index fd8b51c..1253216 100644
--- a/components/SmartChatBox/SmartChatBox.tsx
+++ b/components/SmartChatBox/SmartChatBox.tsx
@@ -16,6 +16,7 @@ interface SmartChatBoxProps {
   previewVideoUrl?: string | null;
   previewVideoId?: string | null;
   onClearPreview?: () => void;
+  setIsFocusChatInput?: (v: boolean) => void;
 }
 
 interface MessageGroup {
@@ -42,7 +43,8 @@ export default function SmartChatBox({
   userId,
   previewVideoUrl,
   previewVideoId,
-  onClearPreview
+  onClearPreview,
+  setIsFocusChatInput
 }: SmartChatBoxProps) {
   // 消息列表引用
   const listRef = useRef<HTMLDivElement>(null);
@@ -197,6 +199,7 @@ export default function SmartChatBox({
         }}
         initialVideoUrl={previewVideoUrl || undefined}
         initialVideoId={previewVideoId || undefined}
+        setIsFocusChatInput={setIsFocusChatInput}
       />
     </div>
   );
diff --git a/components/SmartChatBox/api.ts b/components/SmartChatBox/api.ts
index 76e291b..c24b255 100644
--- a/components/SmartChatBox/api.ts
+++ b/components/SmartChatBox/api.ts
@@ -294,7 +294,7 @@ function transformSystemMessage(
     case 'generate_script_summary':
       if (isScriptSummary(customData)) {
         blocks = [
-          { type: 'text',  text: `🎬 剧本摘要生成完成\n\n${customData.summary}\n${content}` }
+          { type: 'text',  text: `🎬 剧本摘要生成完成\n\n${customData.summary}\n\n${content}` }
         ];
       }
       break;
diff --git a/components/pages/work-flow.tsx b/components/pages/work-flow.tsx
index 84b6421..ca04305 100644
--- a/components/pages/work-flow.tsx
+++ b/components/pages/work-flow.tsx
@@ -28,6 +28,7 @@ const WorkFlow = React.memo(function WorkFlow() {
   const [isSmartChatBoxOpen, setIsSmartChatBoxOpen] = React.useState(true);
   const [previewVideoUrl, setPreviewVideoUrl] = React.useState<string | null>(null);
   const [previewVideoId, setPreviewVideoId] = React.useState<string | null>(null);
+  const [isFocusChatInput, setIsFocusChatInput] = React.useState(false);
 
   const searchParams = useSearchParams();
   const episodeId = searchParams.get('episodeId') || '';
@@ -160,7 +161,7 @@ const WorkFlow = React.memo(function WorkFlow() {
               {taskObject.currentStage !== 'final_video' && taskObject.currentStage !== 'script' && (
                 <div className="h-[123px] w-[calc((100vh-6rem-200px)/9*16)]">
                   <ThumbnailGrid
-                    isDisabledFocus={isEditModalOpen || isPauseWorkFlow || isSmartChatBoxOpen}
+                    isDisabledFocus={isEditModalOpen || isPauseWorkFlow || isFocusChatInput}
                     taskObject={taskObject}
                     currentSketchIndex={currentSketchIndex}
                     onSketchSelect={setCurrentSketchIndex}
@@ -214,6 +215,7 @@ const WorkFlow = React.memo(function WorkFlow() {
           autoFocus={false}
           mask={false}
           zIndex={49}
+          rootClassName="outline-none"
           className="backdrop-blur-lg bg-black/30 border border-white/20 shadow-xl"
           style={{
             backgroundColor: 'transparent',
@@ -236,6 +238,7 @@ const WorkFlow = React.memo(function WorkFlow() {
             userId={userId}
             previewVideoUrl={previewVideoUrl}
             previewVideoId={previewVideoId}
+            setIsFocusChatInput={setIsFocusChatInput}
             onClearPreview={() => {
               setPreviewVideoUrl(null);
               setPreviewVideoId(null);
diff --git a/components/pages/work-flow/thumbnail-grid.tsx b/components/pages/work-flow/thumbnail-grid.tsx
index b089323..8e2e035 100644
--- a/components/pages/work-flow/thumbnail-grid.tsx
+++ b/components/pages/work-flow/thumbnail-grid.tsx
@@ -117,20 +117,12 @@ export function ThumbnailGrid({
   useEffect(() => {
     // 组件挂载时自动聚焦
     if (thumbnailsRef.current && !isDisabledFocus) {
-      thumbnailsRef.current.focus();
       window.addEventListener('keydown', handleKeyDown);
     }
 
     return () => window.removeEventListener('keydown', handleKeyDown);
   }, [handleKeyDown, isDisabledFocus]);
 
-  // 确保在数据变化时保持焦点
-  useEffect(() => {
-    if (thumbnailsRef.current && !isFocused && !isDisabledFocus) {
-      thumbnailsRef.current.focus();
-    }
-  }, [taskObject.currentStage, isFocused]);
-
   // 处理鼠标/触摸拖动事件
   const handleMouseDown = (e: React.MouseEvent<HTMLDivElement>) => {
     // 阻止默认的拖拽行为