diff --git a/app/(admin)/settings/video-canvas/video-canvas-settings-client.tsx b/app/(admin)/settings/video-canvas/video-canvas-settings-client.tsx
index f4282fe52..f9743b16d 100644
--- a/app/(admin)/settings/video-canvas/video-canvas-settings-client.tsx
+++ b/app/(admin)/settings/video-canvas/video-canvas-settings-client.tsx
@@ -175,68 +175,77 @@ export function VideoCanvasSettingsClient() {
         </CardContent>
       </Card>
 
-      <div className="grid gap-4">
-        {functions.map((item) => (
-          <Card className={!item.enabled ? "opacity-70" : undefined} key={item.key}>
-            <CardHeader className="space-y-4">
-              <div className="flex flex-col gap-3 md:flex-row md:items-start md:justify-between">
-                <div>
-                  <CardTitle className="flex flex-wrap items-center gap-2 text-base">
-                    <span>{item.label}</span>
-                    <span className="rounded-full bg-muted px-2 py-0.5 text-xs font-normal text-muted-foreground">
-                      {functionKindLabels[item.kind]}
-                    </span>
-                    <span className="rounded-full bg-muted px-2 py-0.5 text-xs font-normal text-muted-foreground">
-                      {item.route}
-                    </span>
-                  </CardTitle>
-                  <CardDescription className="mt-2">{item.description}</CardDescription>
+      <section className="space-y-4">
+        <div>
+          <h2 className="text-lg font-semibold tracking-tight">功能预设提示词</h2>
+          <p className="mt-1 max-w-3xl text-sm text-muted-foreground">
+            视频画布运行时会通过 /api/video-canvas/config 读取这里保存的提示词。比如“多机位九宫格”的分镜规则、黑色分割线要求和镜头说明，都由管理员在这里维护，前端不再写死覆盖。
+          </p>
+        </div>
+
+        <div className="grid gap-4">
+          {functions.map((item) => (
+            <Card className={!item.enabled ? "opacity-70" : undefined} key={item.key}>
+              <CardHeader className="space-y-4">
+                <div className="flex flex-col gap-3 md:flex-row md:items-start md:justify-between">
+                  <div>
+                    <CardTitle className="flex flex-wrap items-center gap-2 text-base">
+                      <span>{item.label}</span>
+                      <span className="rounded-full bg-muted px-2 py-0.5 text-xs font-normal text-muted-foreground">
+                        {functionKindLabels[item.kind]}
+                      </span>
+                      <span className="rounded-full bg-muted px-2 py-0.5 text-xs font-normal text-muted-foreground">
+                        {item.route}
+                      </span>
+                    </CardTitle>
+                    <CardDescription className="mt-2">{item.description}</CardDescription>
+                  </div>
+                  <div className="flex items-center gap-3">
+                    <Label htmlFor={`enabled-${item.key}`}>启用</Label>
+                    <Switch
+                      checked={item.enabled}
+                      id={`enabled-${item.key}`}
+                      onCheckedChange={(checked) => updateFunction(item.key, { enabled: checked })}
+                    />
+                  </div>
                 </div>
-                <div className="flex items-center gap-3">
-                  <Label htmlFor={`enabled-${item.key}`}>启用</Label>
-                  <Switch
-                    checked={item.enabled}
-                    id={`enabled-${item.key}`}
-                    onCheckedChange={(checked) => updateFunction(item.key, { enabled: checked })}
-                  />
+                <div className="grid gap-4 md:grid-cols-[1fr_120px]">
+                  <div className="space-y-2">
+                    <Label htmlFor={`model-${item.key}`}>模型覆盖</Label>
+                    <Input
+                      id={`model-${item.key}`}
+                      onChange={(event) => updateFunction(item.key, { model: event.target.value })}
+                      placeholder="留空使用全局默认模型"
+                      value={item.model || ""}
+                    />
+                  </div>
+                  <div className="space-y-2">
+                    <Label htmlFor={`credit-${item.key}`}>扣点</Label>
+                    <Input
+                      id={`credit-${item.key}`}
+                      min={0}
+                      onChange={(event) =>
+                        updateFunction(item.key, { creditCost: Number(event.target.value || 0) })
+                      }
+                      type="number"
+                      value={item.creditCost ?? 0}
+                    />
+                  </div>
                 </div>
-              </div>
-              <div className="grid gap-4 md:grid-cols-[1fr_120px]">
-                <div className="space-y-2">
-                  <Label htmlFor={`model-${item.key}`}>模型覆盖</Label>
-                  <Input
-                    id={`model-${item.key}`}
-                    onChange={(event) => updateFunction(item.key, { model: event.target.value })}
-                    placeholder="留空使用全局默认模型"
-                    value={item.model || ""}
-                  />
-                </div>
-                <div className="space-y-2">
-                  <Label htmlFor={`credit-${item.key}`}>扣点</Label>
-                  <Input
-                    id={`credit-${item.key}`}
-                    min={0}
-                    onChange={(event) =>
-                      updateFunction(item.key, { creditCost: Number(event.target.value || 0) })
-                    }
-                    type="number"
-                    value={item.creditCost ?? 0}
-                  />
-                </div>
-              </div>
-            </CardHeader>
-            <CardContent>
-              <Label htmlFor={`prompt-${item.key}`}>预设提示词 / 规则</Label>
-              <Textarea
-                className="mt-2 min-h-28"
-                id={`prompt-${item.key}`}
-                onChange={(event) => updateFunction(item.key, { prompt: event.target.value })}
-                value={item.prompt}
-              />
-            </CardContent>
-          </Card>
-        ))}
-      </div>
+              </CardHeader>
+              <CardContent>
+                <Label htmlFor={`prompt-${item.key}`}>管理员可编辑的预设提示词 / 后端规则</Label>
+                <Textarea
+                  className="mt-2 min-h-28"
+                  id={`prompt-${item.key}`}
+                  onChange={(event) => updateFunction(item.key, { prompt: event.target.value })}
+                  value={item.prompt}
+                />
+              </CardContent>
+            </Card>
+          ))}
+        </div>
+      </section>
     </main>
   );
 }
diff --git a/components/canvas/LiblibVideoCanvasReplica.tsx b/components/canvas/LiblibVideoCanvasReplica.tsx
index 7e0c18ea9..0c8616c22 100644
--- a/components/canvas/LiblibVideoCanvasReplica.tsx
+++ b/components/canvas/LiblibVideoCanvasReplica.tsx
@@ -387,13 +387,10 @@ const OFFICIAL_ZOOM_DOCK_SCALE = 0.65;
 const OFFICIAL_DROPDOWN_SCALE = 0.7;
 const OFFICIAL_NINE_GRID_WIDTH = 930;
 const OFFICIAL_NINE_GRID_HEIGHT = 520;
+const MULTI_VIEW_NINE_GRID_FUNCTION_KEY = "multi-view-nine-grid";
 const PROMPT_COMPOSER_BASE_WIDTH = 895;
 const PROMPT_COMPOSER_BASE_HEIGHT = 286;
 const PRESET_COMPOSER_BASE_HEIGHT = 238;
-const LEGACY_MULTI_VIEW_NINE_GRID_PROMPT =
-  "以选中图片为角色/主体参考，生成多视角 3x3 九宫格分镜图。保持主体一致，九个镜头包含远景、全身、侧身、背面、半身、近景、局部细节等视角，使用清晰黑色分割线分隔每一格，白色或干净背景。";
-const MULTI_VIEW_NINE_GRID_PROMPT =
-  "以选中图片作为角色、服装、发型、场景光线的唯一参考，生成一张 3x3 多机位分镜合成图。必须是 9 个明显不同的镜头画面，不是同一张图裁切或重复人脸。九格分别包含：1 森林环境远景/主体很小，2 全身远景，3 中景正面，4 侧面半身，5 低机位或俯拍，6 面部近景，7 手部/衣袖/动作细节，8 背面或侧背，9 仰视或情绪特写。保持同一角色身份、绿色服装、自然森林光线一致，每格构图、景别、机位和主体大小必须不同。输出为单张 16:9 图片，九格之间使用清晰黑色分割线，不能出现文字、水印、编号，不能白底棚拍。";
 const NINE_GRID_CELL_VISUALS: GridCellVisual[] = [
   { fit: "contain", objectPosition: "50% 50%", scale: 0.62 },
   { fit: "contain", objectPosition: "58% 50%", scale: 0.74, translateX: -10 },
@@ -1016,6 +1013,7 @@ function createMediaResourceNode({
 function createNineGridCompositeNode({
   anchor,
   columns,
+  gridPrompt,
   option,
   rows,
   sourceUrl,
@@ -1024,6 +1022,7 @@ function createNineGridCompositeNode({
 }: {
   anchor: CloneNode;
   columns: number;
+  gridPrompt?: string;
   option?: string;
   rows: number;
   sourceUrl: string;
@@ -1035,10 +1034,11 @@ function createNineGridCompositeNode({
     cellCount === 9
       ? "多机位九宫格"
       : option?.trim() || `${cellCount}宫格`;
-  const gridPrompt =
-    cellCount === 9 && (!option || /多|9|九/.test(option))
-      ? MULTI_VIEW_NINE_GRID_PROMPT
-      : `基于选中图片生成 ${columns}x${rows} 宫格分镜图，使用黑色分割线分隔每一格。`;
+  const resolvedGridPrompt =
+    gridPrompt?.trim() ||
+    (cellCount === 9 && (!option || /多|9|九/.test(option))
+      ? getDefaultFunctionPrompt(MULTI_VIEW_NINE_GRID_FUNCTION_KEY)
+      : `基于选中图片生成 ${columns}x${rows} 宫格分镜图，使用黑色分割线分隔每一格。`);
 
   return {
     id: makeId("image"),
@@ -1055,7 +1055,7 @@ function createNineGridCompositeNode({
       action: "IMAGE_GENERATE_NINE_GRID",
       generated: true,
       gridColumns: columns,
-      gridPrompt,
+      gridPrompt: resolvedGridPrompt,
       gridRows: rows,
       mediaType: "image",
       nineGridCells: Array.from({ length: cellCount }).map((_, index) => getGridCellVisual(index)),
@@ -1066,7 +1066,7 @@ function createNineGridCompositeNode({
         columns,
         mode: "grid",
         option: option || "",
-        prompt: gridPrompt,
+        prompt: resolvedGridPrompt,
         rows,
         sourceNodeId: anchor.id,
         sourceUrl,
@@ -1174,6 +1174,10 @@ function getConfiguredFunction(
   return config.functions[key] || defaultVideoCanvasFunctionConfig.functions[key];
 }
 
+function getDefaultFunctionPrompt(key: string) {
+  return defaultVideoCanvasFunctionConfig.functions[key]?.prompt || "";
+}
+
 function getFunctionModel(
   config: VideoCanvasFunctionConfig,
   definition: VideoCanvasFunctionDefinition,
@@ -1203,16 +1207,10 @@ function composeMultiViewNineGridPrompt(
   definition: VideoCanvasFunctionDefinition,
   option?: string,
 ) {
-  const configuredPrompt = definition.prompt.trim();
+  const configuredPrompt = definition.prompt.trim() || getDefaultFunctionPrompt(MULTI_VIEW_NINE_GRID_FUNCTION_KEY);
   const selectedOption = String(option || "").trim();
-  const parts = [MULTI_VIEW_NINE_GRID_PROMPT];
-  const isLegacyPrompt =
-    configuredPrompt === LEGACY_MULTI_VIEW_NINE_GRID_PROMPT ||
-    (configuredPrompt.includes("以选中图片为角色/主体参考") && configuredPrompt.includes("白色或干净背景"));
+  const parts = [configuredPrompt];
 
-  if (configuredPrompt && configuredPrompt !== MULTI_VIEW_NINE_GRID_PROMPT && !isLegacyPrompt) {
-    parts.push(`后台补充：${configuredPrompt}`);
-  }
   if (selectedOption) {
     parts.push(`用户选择：${selectedOption}`);
   }
@@ -1544,7 +1542,7 @@ function normalizeRestoredNode(node: CloneNode, fallbackIndex: number): CloneNod
       draftValues: {
         ...(node.draftValues || {}),
         action: "IMAGE_GENERATE_NINE_GRID",
-        gridPrompt: getDraftString(node, "gridPrompt") || MULTI_VIEW_NINE_GRID_PROMPT,
+        gridPrompt: getDraftString(node, "gridPrompt") || getDefaultFunctionPrompt(MULTI_VIEW_NINE_GRID_FUNCTION_KEY),
         nineGridComposite: true,
         outputSizeLabel: getDraftString(node, "outputSizeLabel") || "2752 × 1536",
       },
@@ -2578,10 +2576,12 @@ export function LiblibVideoCanvasReplica({ canvasId, isLoading = false }: Liblib
       const title = labels[action];
       if (action === "grid") {
         const { columns, rows } = normalizeGridOption(option || "9宫格");
-        const gridDefinition = getConfiguredFunction(videoCanvasFunctionConfig, "multi-view-nine-grid");
+        const gridDefinition = getConfiguredFunction(videoCanvasFunctionConfig, MULTI_VIEW_NINE_GRID_FUNCTION_KEY);
+        const prompt = composeMultiViewNineGridPrompt(gridDefinition, option);
         const gridNode = createNineGridCompositeNode({
           anchor,
           columns,
+          gridPrompt: prompt,
           option,
           rows,
           sourceUrl,
@@ -2613,7 +2613,6 @@ export function LiblibVideoCanvasReplica({ canvasId, isLoading = false }: Liblib
         setActivePanel(null);
 
         if (videoCanvasFunctionConfig.enabled && gridDefinition?.enabled) {
-          const prompt = composeMultiViewNineGridPrompt(gridDefinition, option);
           const response = await imageGenerationService.generateImage({
             aspectRatio: "16:9",
             imageCount: 1,
@@ -2852,7 +2851,7 @@ export function LiblibVideoCanvasReplica({ canvasId, isLoading = false }: Liblib
       const starterMode = getPresetMode(anchor);
       const functionKey =
         payload.slashScene && /九宫格|宫格|多机位/.test(payload.slashScene)
-          ? "multi-view-nine-grid"
+          ? MULTI_VIEW_NINE_GRID_FUNCTION_KEY
           : starterMode === "character-three-view"
             ? "character-three-view"
             : "image-repaint";
@@ -2862,7 +2861,7 @@ export function LiblibVideoCanvasReplica({ canvasId, isLoading = false }: Liblib
         payload.prompt.trim() || payload.slashScene || "生成角色三视图",
       );
       const finalPrompt =
-        functionKey === "multi-view-nine-grid"
+        functionKey === MULTI_VIEW_NINE_GRID_FUNCTION_KEY
           ? composeMultiViewNineGridPrompt(definition, payload.prompt.trim() || payload.slashScene || "多机位九宫格")
           : prompt;
       const count = Math.max(1, Math.min(4, Number.parseInt(payload.settings.count, 10) || 1));
@@ -2885,14 +2884,14 @@ export function LiblibVideoCanvasReplica({ canvasId, isLoading = false }: Liblib
       const generatedNode = generatedUrl
         ? createMediaResourceNode({
             action: "IMAGE_RESOURCE",
-            height: functionKey === "multi-view-nine-grid" ? OFFICIAL_NINE_GRID_HEIGHT : 280,
+            height: functionKey === MULTI_VIEW_NINE_GRID_FUNCTION_KEY ? OFFICIAL_NINE_GRID_HEIGHT : 280,
             mediaType: "image",
             mediaUrl: generatedUrl,
             prompt: finalPrompt,
             title: payload.slashScene || definition.label || "图片生成",
-            width: functionKey === "multi-view-nine-grid" ? OFFICIAL_NINE_GRID_WIDTH : 420,
+            width: functionKey === MULTI_VIEW_NINE_GRID_FUNCTION_KEY ? OFFICIAL_NINE_GRID_WIDTH : 420,
             x: rightX,
-            y: Math.round(anchor.y + anchor.height / 2 - (functionKey === "multi-view-nine-grid" ? OFFICIAL_NINE_GRID_HEIGHT : 280) / 2),
+            y: Math.round(anchor.y + anchor.height / 2 - (functionKey === MULTI_VIEW_NINE_GRID_FUNCTION_KEY ? OFFICIAL_NINE_GRID_HEIGHT : 280) / 2),
           })
         : null;
       const created = generatedNode || createTargetModuleNode("image", rightX, 0, nodes.length + 1);
@@ -3118,13 +3117,16 @@ export function LiblibVideoCanvasReplica({ canvasId, isLoading = false }: Liblib
       if (node.inputType === "image" || /IMAGE_GENERATE/.test(action)) {
         const functionKey =
           action === "IMAGE_GENERATE_NINE_GRID"
-            ? "multi-view-nine-grid"
+            ? MULTI_VIEW_NINE_GRID_FUNCTION_KEY
             : node.draftValues?.generatorType === "ENHANCE"
               ? "hd-upscale"
               : "image-repaint";
         const definition = getConfiguredFunction(videoCanvasFunctionConfig, functionKey);
         const sourceUrl = imageUrls[0] || getNodeSourceMediaUrl(node);
-        const prompt = composeFunctionPrompt(definition, rawPrompt || String(params.option || node.title));
+        const prompt =
+          functionKey === MULTI_VIEW_NINE_GRID_FUNCTION_KEY
+            ? composeMultiViewNineGridPrompt(definition, String(params.option || node.title))
+            : composeFunctionPrompt(definition, rawPrompt || String(params.option || node.title));
         let generatedUrl = "";
 
         if (definition.route === "enhance" && sourceUrl) {