diff --git a/components/SmartChatBox/InputBar.tsx b/components/SmartChatBox/InputBar.tsx
index cf8b1b5..ea3cde8 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);
@@ -91,6 +92,22 @@ export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVide
     debouncedAdjustHeight();
   }, [text, debouncedAdjustHeight]);
 
+  // 布局切换时保持输入框焦点并将光标移到末尾
+  useEffect(() => {
+    // 等待布局动画完成后再聚焦，避免动画过程中的视觉跳动
+    const focusTimeout = setTimeout(() => {
+      const textarea = textareaRef.current;
+      if (textarea) {
+        textarea.focus();
+        // 将光标移动到文本末尾
+        const length = textarea.value.length;
+        textarea.setSelectionRange(length, length);
+      }
+    }, 200); // 与布局动画时长保持一致
+
+    return () => clearTimeout(focusTimeout);
+  }, [isMultiline]);
+
   const handleSend = () => {
     const blocks: MessageBlock[] = [];
     if (text.trim()) blocks.push({ type: "text" as const, text: text.trim() });
@@ -229,6 +246,7 @@ export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVide
             {/* 图片上传按钮 - 单行时显示在左侧 */}
             {!isMultiline && (
               <motion.label 
+                key="single-line-upload"
                 initial={{ opacity: 0, scale: 0.8 }}
                 animate={{ opacity: 1, scale: 1 }}
                 exit={{ opacity: 0, scale: 0.8 }}
@@ -247,11 +265,11 @@ export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVide
             )}
 
             {/* 文本输入 */}
-            <motion.div layout className="flex-1">
+            <motion.div layout key="text-input" className="flex-1 flex">
               <textarea
                 ref={textareaRef}
                 placeholder="输入文字…"
-                className="w-full pl-[10px] pr-[10px] py-[1rem] rounded-[10px] leading-[20px] text-sm border-none bg-transparent text-white placeholder:text-white/[0.40] focus:outline-none resize-none min-h-[48px] max-h-[120px] overflow-y-auto"
+                className="w-full pl-2 pr-2 py-4 rounded-2 leading-4 text-sm border-none bg-transparent text-white placeholder:text-white/[0.40] focus:outline-none resize-none min-h-[48px] max-h-[120px] overflow-y-auto"
                 rows={1}
                 value={text}
                 onChange={(e) => setText(e.target.value)}
@@ -269,6 +287,8 @@ export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVide
                     handleSend();
                   }
                 }}
+                onFocus={() => setIsFocusChatInput?.(true)}
+                onBlur={() => setIsFocusChatInput?.(false)}
                 data-alt="text-input"
               />
             </motion.div>
@@ -276,6 +296,7 @@ export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVide
             {isMultiline ? (
               // 多行模式：底部按钮区域
               <motion.div 
+                key="multi-line-buttons"
                 layout
                 initial={{ opacity: 0, y: -10 }}
                 animate={{ opacity: 1, y: 0 }}
@@ -284,6 +305,7 @@ export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVide
               >
                 {/* 图片上传 */}
                 <motion.label 
+                  key="multi-line-upload"
                   whileHover={{ scale: 1.05 }}
                   whileTap={{ scale: 0.95 }}
                   className={`cursor-pointer inline-flex items-center gap-2 p-2 rounded-full hover:bg-gray-700/50 text-gray-100 ${isUploading ? 'opacity-50 cursor-not-allowed' : ''}`} 
@@ -314,6 +336,7 @@ export function InputBar({ onSend, setVideoPreview, initialVideoUrl, initialVide
             ) : (
               // 单行模式：发送按钮
               <motion.button
+                key="single-line-send"
                 whileHover={{ scale: 1.05 }}
                 whileTap={{ scale: 0.95 }}
                 onClick={handleSend}
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