diff --git a/src/app/api/generate/providers/__tests__/popiserver.test.ts b/src/app/api/generate/providers/__tests__/popiserver.test.ts
index 4ce538ab..0091e327 100644
--- a/src/app/api/generate/providers/__tests__/popiserver.test.ts
+++ b/src/app/api/generate/providers/__tests__/popiserver.test.ts
@@ -106,10 +106,10 @@ describe("popiserver generation provider", () => {
       aiModelCode: "viduq2-pro",
       aiModelCodeAlias: "viduq2-pro",
       ratio: "16:9",
+      aspectRatio: "16:9",
       resolution: "720",
       videoRatio: 10,
     });
-    expect(body.aspectRatio).toBeUndefined();
   });
 
   it("returns the prepared PopiServer task request when gateway skip is enabled", async () => {
@@ -316,8 +316,8 @@ describe("popiserver generation provider", () => {
     const body = JSON.parse(String(init?.body));
     expect(body).toMatchObject({
       aspect_ratio: "9:16",
+      aspectRatio: "9:16",
     });
-    expect(body.aspectRatio).toBeUndefined();
   });
 
   it("embeds ratio and resolution in chatPrompt for gpt-image-2-vip", async () => {
@@ -357,6 +357,7 @@ describe("popiserver generation provider", () => {
     expect(body).toMatchObject({
       chatPrompt: "比例：16:9\n清晰度：1K\n提示词：create a cinematic cat",
       ratio: "16:9",
+      aspectRatio: "16:9",
       resolution: "1K",
     });
   });
@@ -398,9 +399,9 @@ describe("popiserver generation provider", () => {
     expect(body).toMatchObject({
       chatPrompt: "比例：9:16\n清晰度：2K\n提示词：create a clean product render",
       ratio: "9:16",
+      aspectRatio: "9:16",
       resolution: "2K",
     });
-    expect(body.aspectRatio).toBeUndefined();
   });
 
   it("uses frontend referenceSubjectList for image tasks", async () => {
diff --git a/src/components/WorkflowCanvas.tsx b/src/components/WorkflowCanvas.tsx
index e42b35c5..e56cdb46 100644
--- a/src/components/WorkflowCanvas.tsx
+++ b/src/components/WorkflowCanvas.tsx
@@ -418,38 +418,6 @@ const getNodeDimensions = (node: Node): { width: number; height: number } => {
   };
 };
 
-const hasAutoSizedMediaContent = (node: Node): boolean => {
-  const data = (node.data ?? {}) as Record<string, unknown>;
-  switch (node.type) {
-    case "imageInput":
-    case "smartImage":
-      return Boolean(data.image || data.outputImage);
-    case "videoInput":
-    case "smartVideo":
-      return Boolean(data.video || data.previewVideoPoster || data.outputVideo);
-    case "nanoBanana":
-      return Boolean(data.outputImage);
-    case "generateVideo":
-      return Boolean(data.outputVideo);
-    default:
-      return false;
-  }
-};
-
-const stripNodeSizeForAutoMedia = (node: Node): Node => {
-  if (!hasAutoSizedMediaContent(node)) return node;
-  const style = node.style as Record<string, unknown> | undefined;
-  if (!style && node.width === undefined && node.height === undefined) return node;
-
-  const { width: _width, height: _height, ...restStyle } = style ?? {};
-  return {
-    ...node,
-    width: undefined,
-    height: undefined,
-    style: Object.keys(restStyle).length > 0 ? restStyle : undefined,
-  };
-};
-
 const QUICK_ADD_AFFORDANCE_OFFSET_X = 18;
 const QUICK_ADD_HOT_ZONE_WIDTH = 96;
 const QUICK_ADD_HOT_ZONE_HEIGHT = 96;
@@ -1432,7 +1400,7 @@ export function WorkflowCanvas() {
   });
 
   const reactFlowNodes = useMemo(
-    () => [...allNodes.map(stripNodeSizeForAutoMedia), ...batchNodes],
+    () => [...allNodes, ...batchNodes],
     [allNodes, batchNodes]
   );
 
diff --git a/src/utils/__tests__/generationConfig.test.ts b/src/utils/__tests__/generationConfig.test.ts
index 27e4e4e7..3ec94d44 100644
--- a/src/utils/__tests__/generationConfig.test.ts
+++ b/src/utils/__tests__/generationConfig.test.ts
@@ -421,6 +421,7 @@ describe("generationConfig", () => {
         subType: POPI_VIDEO_SUBTYPE_IMAGE_TO_VIDEO,
         referenceSubjectList: [],
         parameters: {
+          ratio: "16:9",
           resolution: "720p",
           duration: 4,
           videoRatio: 20,
@@ -440,6 +441,8 @@ describe("generationConfig", () => {
       chatPrompt: "prompt",
       origin: "canvas",
       aiModelName: "Popi Video",
+      ratio: "16:9",
+      aspectRatio: "16:9",
       resolution: "720p",
       duration: 4,
       videoRatio: 20,
diff --git a/src/utils/popiserverTaskPayload.ts b/src/utils/popiserverTaskPayload.ts
index 402f6df2..8c880155 100644
--- a/src/utils/popiserverTaskPayload.ts
+++ b/src/utils/popiserverTaskPayload.ts
@@ -64,6 +64,12 @@ function resolveAspectRatio(parameters: Record<string, unknown> | undefined): st
   return asString(parameters?.ratio) ?? asString(parameters?.aspectRatio) ?? asString(parameters?.aspect_ratio) ?? "";
 }
 
+function mapRatioToAspectRatio(parameters: Record<string, unknown>): Record<string, unknown> {
+  if (asString(parameters.aspectRatio)) return parameters;
+  const aspectRatio = asString(parameters.ratio) ?? asString(parameters.aspect_ratio);
+  return aspectRatio ? { ...parameters, aspectRatio } : parameters;
+}
+
 function needsPromptEmbeddedParameters(modelCode: string, modelAlias: string): boolean {
   return (
     modelCode === "gpt-image-2-vip" ||
@@ -105,7 +111,7 @@ export function buildPopiTaskBody(
   const rawParameters = removeLegacyAudioParameters(input.parameters || {});
   const rawExtraTaskParams = removeLegacyAudioParameters(input.extraTaskParams || {});
   const aspectRatio = resolveAspectRatio(rawParameters);
-  const parameters = rawParameters;
+  const parameters = mapRatioToAspectRatio(rawParameters);
   const selectedModel = {
     provider: "popiserver" as const,
     modelId: input.model.id,