修改音频 视频问题

src/components/__tests__/GenerationComposer.test.tsx

@@ -1692,7 +1692,7 @@ describe("GenerationComposer", () => {
     });
     render(<GenerationComposer />);
 
-    fireEvent.click(screen.getByText("First/last frame"));
+    fireEvent.click(screen.getByText("首尾帧"));
 
     await waitFor(() => {
       const data = useWorkflowStore.getState().nodes[3].data as GenerateVideoNodeData;
@@ -1714,7 +1714,7 @@ describe("GenerationComposer", () => {
     });
     render(<GenerationComposer />);
 
-    fireEvent.click(screen.getByText("Image to video"));
+    fireEvent.click(screen.getByText("图生视频"));
 
     await waitFor(() => {
       const data = useWorkflowStore.getState().nodes[1].data as GenerateVideoNodeData;
@@ -1738,7 +1738,38 @@ describe("GenerationComposer", () => {
     });
     render(<GenerationComposer />);
 
-    fireEvent.click(screen.getByText("Image to video"));
+    fireEvent.click(screen.getByText("图生视频"));
+
+    await waitFor(() => {
+      const data = useWorkflowStore.getState().nodes[2].data as GenerateVideoNodeData;
+      expect(data.parameters?.subType).toBe(203);
+    });
+  });
+
+  it("does not allow image-to-video when video material is connected", async () => {
+    useWorkflowStore.setState({
+      nodes: [
+        imageNode("img-a", false, { outputImage: "https://static.popi.art/a.jpg" }),
+        videoNode("source-video", false, { outputVideo: "https://static.popi.art/source.mp4" }),
+        videoNode("vid-1", true, {
+          parameters: { subType: 203 },
+        }),
+      ],
+      edges: [
+        imageEdge("img-a", "vid-1", 0),
+        {
+          id: "edge-source-video-vid-1-output-input",
+          source: "source-video",
+          sourceHandle: "output",
+          target: "vid-1",
+          targetHandle: "input",
+          data: { createdAt: 1 },
+        } as WorkflowEdge,
+      ],
+    });
+    render(<GenerationComposer />);
+
+    fireEvent.click(screen.getByText("图生视频"));
 
     await waitFor(() => {
       const data = useWorkflowStore.getState().nodes[2].data as GenerateVideoNodeData;
@@ -1768,6 +1799,37 @@ describe("GenerationComposer", () => {
     });
   });
 
+  it("falls back from first/last frame when audio material is connected", async () => {
+    useWorkflowStore.setState({
+      nodes: [
+        imageNode("img-a", false, { outputImage: "https://static.popi.art/a.jpg" }),
+        imageNode("img-b", false, { outputImage: "https://static.popi.art/b.jpg" }),
+        audioNode("source-audio", false, { outputAudio: "https://static.popi.art/source.mp3" }),
+        videoNode("vid-1", true, {
+          parameters: { subType: 204 },
+        }),
+      ],
+      edges: [
+        imageEdge("img-a", "vid-1", 0),
+        imageEdge("img-b", "vid-1", 1),
+        {
+          id: "edge-source-audio-vid-1-output-input",
+          source: "source-audio",
+          sourceHandle: "output",
+          target: "vid-1",
+          targetHandle: "input",
+          data: { createdAt: 2 },
+        } as WorkflowEdge,
+      ],
+    });
+    render(<GenerationComposer />);
+
+    await waitFor(() => {
+      const data = useWorkflowStore.getState().nodes[3].data as GenerateVideoNodeData;
+      expect(data.parameters?.subType).toBe(203);
+    });
+  });
+
   it("labels two first/last frame images without changing their reference order", async () => {
     useWorkflowStore.setState({
       nodes: [
@@ -1785,10 +1847,10 @@ describe("GenerationComposer", () => {
     });
     render(<GenerationComposer />);
 
-    expect(screen.getByText("First")).toBeInTheDocument();
-    expect(screen.getByText("Last")).toBeInTheDocument();
+    expect(screen.getByText("首帧")).toBeInTheDocument();
+    expect(screen.getByText("尾帧")).toBeInTheDocument();
 
-    fireEvent.click(screen.getByLabelText("Generate"));
+    fireEvent.click(screen.getByLabelText(/Generate|生成/));
 
     await waitFor(() => {
       expect(useWorkflowStore.getState().regenerateNode).toHaveBeenCalledWith("vid-1");
@@ -1799,13 +1861,13 @@ describe("GenerationComposer", () => {
         id: 1,
         type: "image",
         url: "https://static.popi.art/first.jpg",
-        name: "@图1",
+        name: "图1",
       },
       {
         id: 2,
         type: "image",
         url: "https://static.popi.art/last.jpg",
-        name: "@图2",
+        name: "图2",
       },
     ]);
   });

src/components/__tests__/ModelParameters.test.tsx

@@ -26,7 +26,7 @@ describe("ModelParameters", () => {
     const { rerender } = render(
       <ModelParameters {...defaultProps} isLoading schema={[]} />
     );
-    expect(screen.getByText("Loading parameters...")).toBeInTheDocument();
+    expect(screen.getByText(/Loading parameters|正在加载参数/)).toBeInTheDocument();
 
     rerender(
       <ModelParameters {...defaultProps} error="Failed to load model" schema={[]} />
@@ -170,11 +170,52 @@ describe("ModelParameters", () => {
       />
     );
 
-    expect(screen.getByText("Quality")).toBeInTheDocument();
-    expect(screen.queryByText("Size")).not.toBeInTheDocument();
+    expect(screen.getByRole("combobox")).toBeInTheDocument();
+    expect(screen.queryByText("1024 x 1024")).not.toBeInTheDocument();
     expect(onParametersChange).toHaveBeenCalledWith({});
     await waitFor(() => {
       expect(onParametersChange).toHaveBeenCalledWith({ quality: "auto" });
     });
   });
+
+  it("can hide parameters without pruning stored values", async () => {
+    const onParametersChange = vi.fn();
+    render(
+      <ModelParameters
+        {...defaultProps}
+        parameters={{ voiceId: "male-qn-qingse" }}
+        schema={[
+          createMockParameter({
+            name: "voiceId",
+            type: "string",
+            options: [
+              { value: "male-qn-qingse", label: "Male" },
+              { value: "female-shaonv", label: "Female" },
+            ],
+            default: "female-shaonv",
+          }),
+          createMockParameter({
+            name: "quality",
+            type: "string",
+            options: [
+              { value: "auto", label: "Auto" },
+              { value: "high", label: "High" },
+            ],
+            default: "auto",
+          }),
+        ]}
+        onParametersChange={onParametersChange}
+        hiddenParameterNames={["voiceId"]}
+        pruneHiddenParameters={false}
+      />
+    );
+
+    expect(screen.queryByText("VoiceId")).not.toBeInTheDocument();
+    await waitFor(() => {
+      expect(onParametersChange).toHaveBeenCalledWith({
+        voiceId: "male-qn-qingse",
+        quality: "auto",
+      });
+    });
+  });
 });

src/components/composer/GenerationComposer.tsx

@@ -97,6 +97,11 @@ type VideoStitchLoopCount = 1 | 2 | 3;
 type VideoGenerationSubType = 202 | 203 | 204;
 type ComposerReferenceMaterialType = "image" | "video" | "audio";
 type AssistMenuType = "command" | "reference";
+type ReferenceMediaCounts = {
+  images: number;
+  videos: number;
+  audio: number;
+};
 
 interface ComposerReferenceMaterial {
   id: number;
@@ -247,12 +252,12 @@ function videoGenerationSubTypeFromUnknown(value: unknown): VideoGenerationSubTy
   return isVideoGenerationSubType(parsed) ? parsed : null;
 }
 
-function canUseFirstLastFrameSubType(imageCount: number): boolean {
-  return imageCount === 2;
+function canUseFirstLastFrameSubType(counts: ReferenceMediaCounts): boolean {
+  return counts.images === 2 && counts.videos === 0 && counts.audio === 0;
 }
 
-function canUseImageToVideoSubType(imageCount: number): boolean {
-  return imageCount === 1;
+function canUseImageToVideoSubType(counts: ReferenceMediaCounts): boolean {
+  return counts.images === 1 && counts.videos === 0 && counts.audio === 0;
 }
 
 function getSchemaSelectOptions(param: ModelParameter): SchemaSelectOption[] {
@@ -1856,8 +1861,13 @@ export function GenerationComposer() {
   const connectedReferenceAudio = context.mode === "node-edit" ? connectedInputs?.audio ?? [] : [];
   const referenceImages = connectedReferenceImages.length > 0 ? connectedReferenceImages : draft.inputImages;
   const isVideoSubTypeSelectorVisible = context.mode === "node-edit" && context.nodeType === "generateVideo";
-  const canUseFirstLastFrame = canUseFirstLastFrameSubType(referenceImages.length);
-  const canUseImageToVideo = canUseImageToVideoSubType(referenceImages.length);
+  const referenceMediaCounts: ReferenceMediaCounts = {
+    images: referenceImages.length,
+    videos: context.mode === "node-edit" ? connectedInputs?.videos.length ?? 0 : 0,
+    audio: context.mode === "node-edit" ? connectedInputs?.audio.length ?? 0 : 0,
+  };
+  const canUseFirstLastFrame = canUseFirstLastFrameSubType(referenceMediaCounts);
+  const canUseImageToVideo = canUseImageToVideoSubType(referenceMediaCounts);
   const referenceImageInputs = useMemo<ComposerReferenceInput[]>(() => {
     const imageEdgeIds =
       context.mode === "node-edit" && context.node && connectedReferenceImages.length > 0
@@ -2539,6 +2549,11 @@ export function GenerationComposer() {
     const submitReferenceImageCount = connectedInputs?.images.length && connectedInputs.images.length > 0
       ? connectedInputs.images.length
       : activeDraft.inputImages.length;
+    const submitReferenceMediaCounts: ReferenceMediaCounts = {
+      images: submitReferenceImageCount,
+      videos: connectedInputs?.videos.length ?? 0,
+      audio: connectedInputs?.audio.length ?? 0,
+    };
     const referenceSubjectList = buildReferenceSubjectListParameter(referenceMaterials);
     const shouldWriteVideoSubType = context.mode === "empty-create" || explicitSubmitVideoSubType !== null;
     const shouldAttachReferenceSubjectList =
@@ -2563,7 +2578,7 @@ export function GenerationComposer() {
     if (
       submitVideoNodeType === "generateVideo" &&
       explicitSubmitVideoSubType === VIDEO_GENERATION_SUB_TYPE_MULTI_REFERENCE &&
-      !canUseImageToVideoSubType(submitReferenceImageCount)
+      !canUseImageToVideoSubType(submitReferenceMediaCounts)
     ) {
       message.warning(t("composer.videoSubTypeImageToVideoImageCountRequired"));
       return;
@@ -2571,7 +2586,7 @@ export function GenerationComposer() {
     if (
       submitVideoNodeType === "generateVideo" &&
       explicitSubmitVideoSubType === VIDEO_GENERATION_SUB_TYPE_FIRST_LAST_FRAME &&
-      !canUseFirstLastFrameSubType(submitReferenceImageCount)
+      !canUseFirstLastFrameSubType(submitReferenceMediaCounts)
     ) {
       message.warning(t("composer.videoSubTypeFirstLastFrameImageCountRequired"));
       return;

src/components/nodes/GenerateAudioNode.tsx

@@ -495,6 +495,7 @@ export function GenerateAudioNodeView({
                     parameters={nodeConfig.parameters || {}}
                     onParametersChange={handleParametersChange}
                     hiddenParameterNames={AUDIO_COMPOSER_PARAMETER_NAMES}
+                    pruneHiddenParameters={false}
                   />
                 )}
               </>
@@ -714,6 +715,7 @@ export function GenerateAudioNodeView({
             onParametersChange={handleParametersChange}
             onExpandChange={handleParametersExpandChange}
             hiddenParameterNames={AUDIO_COMPOSER_PARAMETER_NAMES}
+            pruneHiddenParameters={false}
           />
         </div>
       )}

src/components/nodes/ModelParameters.tsx

@@ -31,6 +31,7 @@ interface ModelParametersProps {
   onExpandChange?: (expanded: boolean, parameterCount: number) => void;
   onInputsLoaded?: (inputs: ModelInputDef[]) => void;
   hiddenParameterNames?: string[];
+  pruneHiddenParameters?: boolean;
 }
 
 /**
@@ -46,6 +47,7 @@ function ModelParametersInner({
   onParametersChange,
   onExpandChange,
   hiddenParameterNames,
+  pruneHiddenParameters = true,
 }: ModelParametersProps) {
   const { t } = useI18n();
   const lastExpandCountRef = useRef<number | null>(null);
@@ -62,7 +64,7 @@ function ModelParametersInner({
 
   // Remove parameters controlled by a higher-priority surface, such as the bottom composer.
   useEffect(() => {
-    if (hiddenParameterSet.size === 0) return;
+    if (!pruneHiddenParameters || hiddenParameterSet.size === 0) return;
     const nextParameters = { ...parameters };
     let changed = false;
     for (const name of hiddenParameterSet) {
@@ -74,14 +76,16 @@ function ModelParametersInner({
     if (changed) {
       onParametersChange(nextParameters);
     }
-  }, [hiddenParameterSet, onParametersChange, parameters]);
+  }, [hiddenParameterSet, onParametersChange, parameters, pruneHiddenParameters]);
 
   // Pre-populate visible schema defaults into parameters
   useEffect(() => {
     if (visibleSchema.length === 0) return;
     const baseParameters = { ...parameters };
-    for (const name of hiddenParameterSet) {
-      delete baseParameters[name];
+    if (pruneHiddenParameters) {
+      for (const name of hiddenParameterSet) {
+        delete baseParameters[name];
+      }
     }
     const defaults: Record<string, unknown> = {};
     let hasNewDefaults = false;
@@ -101,7 +105,7 @@ function ModelParametersInner({
     if (hasNewDefaults) {
       onParametersChange({ ...baseParameters, ...defaults });
     }
-  }, [visibleSchema, parameters, onParametersChange]);
+  }, [visibleSchema, parameters, onParametersChange, hiddenParameterSet, pruneHiddenParameters]);
 
   // Notify parent to resize node when schema loads
   useEffect(() => {

src/i18n/index.tsx

@@ -708,8 +708,8 @@ const en = {
   "composer.videoSubTypeTextToVideo": "All-purpose reference",
   "composer.videoSubTypeFirstLastFrame": "First/last frame",
   "composer.videoSubTypeImageRequired": "Image to video or first/last frame requires at least one reference image.",
-  "composer.videoSubTypeImageToVideoImageCountRequired": "Image to video requires exactly one reference image.",
-  "composer.videoSubTypeFirstLastFrameImageCountRequired": "First/last frame requires exactly two reference images.",
+  "composer.videoSubTypeImageToVideoImageCountRequired": "Image to video requires exactly one reference image and no audio or video assets.",
+  "composer.videoSubTypeFirstLastFrameImageCountRequired": "First/last frame requires exactly two reference images and no audio or video assets.",
   "composer.firstFrame": "First",
   "composer.lastFrame": "Last",
   "composer.commands": "Commands",
@@ -1854,8 +1854,8 @@ Object.assign(zhCN, {
   "group.colorPurple": "紫色",
   "group.colorOrange": "橙色",
   "group.colorRed": "红色",
-  "composer.videoSubTypeImageToVideoImageCountRequired": "\u56fe\u751f\u89c6\u9891\u9700\u8981\u6070\u597d\u4e00\u5f20\u53c2\u8003\u56fe\u7247\u3002",
-  "composer.videoSubTypeFirstLastFrameImageCountRequired": "\u9996\u5c3e\u5e27\u9700\u8981\u6070\u597d\u4e24\u5f20\u53c2\u8003\u56fe\u7247\u3002",
+  "composer.videoSubTypeImageToVideoImageCountRequired": "\u56fe\u751f\u89c6\u9891\u9700\u8981\u6070\u597d\u4e00\u5f20\u53c2\u8003\u56fe\u7247\uff0c\u4e14\u4e0d\u80fd\u5305\u542b\u97f3\u9891\u6216\u89c6\u9891\u7d20\u6750\u3002",
+  "composer.videoSubTypeFirstLastFrameImageCountRequired": "\u9996\u5c3e\u5e27\u9700\u8981\u6070\u597d\u4e24\u5f20\u53c2\u8003\u56fe\u7247\uff0c\u4e14\u4e0d\u80fd\u5305\u542b\u97f3\u9891\u6216\u89c6\u9891\u7d20\u6750\u3002",
   "composer.firstFrame": "\u9996\u5e27",
   "composer.lastFrame": "\u5c3e\u5e27",
   "composer.videoSubTypeMultiReference": "图生视频",

src/store/execution/__tests__/generateVideoExecutor.test.ts

@@ -68,6 +68,27 @@ beforeEach(() => {
   vi.clearAllMocks();
 });
 
+function mockPopiTaskSuccess() {
+  mockFetch.mockImplementation(async (url: string) => {
+    if (url === "/api/generate/task/create") {
+      return {
+        ok: true,
+        json: () => Promise.resolve({ success: true, taskId: "task-1" }),
+      };
+    }
+    if (url.startsWith("/api/generate/task/poll")) {
+      return {
+        ok: true,
+        json: () => Promise.resolve({
+          status: "completed",
+          result: { success: true, video: "data:video/mp4;base64,output" },
+        }),
+      };
+    }
+    throw new Error(`Unexpected fetch URL: ${url}`);
+  });
+}
+
 async function withProviderMode<T>(mode: "multi" | "popi", run: () => Promise<T>): Promise<T> {
   const originalNextPublicMode = process.env.NEXT_PUBLIC_PROVIDER_MODE;
   const originalProviderMode = process.env.PROVIDER_MODE;
@@ -394,10 +415,7 @@ describe("executeGenerateVideo", () => {
       selectedModel: { provider: "popiserver", modelId: "44", displayName: "Seedance" },
       parameters: { subType: 202 },
     });
-    mockFetch.mockResolvedValueOnce({
-      ok: true,
-      json: () => Promise.resolve({ success: true, video: "data:video/mp4;base64,output" }),
-    });
+    mockPopiTaskSuccess();
 
     const ctx = makeCtx(node, {
       getConnectedInputs: vi.fn().mockReturnValue({
@@ -424,10 +442,7 @@ describe("executeGenerateVideo", () => {
       selectedModel: { provider: "popiserver", modelId: "44", displayName: "Seedance" },
       parameters: { subType: 204 },
     });
-    mockFetch.mockResolvedValueOnce({
-      ok: true,
-      json: () => Promise.resolve({ success: true, video: "data:video/mp4;base64,output" }),
-    });
+    mockPopiTaskSuccess();
 
     const ctx = makeCtx(node, {
       getConnectedInputs: vi.fn().mockReturnValue({
@@ -445,15 +460,68 @@ describe("executeGenerateVideo", () => {
     expect(body.parameters.subType).toBe(204);
   });
 
+  it("normalizes image-to-video subtype to all-purpose reference when video input exists", async () => {
+    const node = makeNode({
+      selectedModel: {
+        provider: "popiserver",
+        modelId: "44",
+        displayName: "Seedance",
+        metadata: { isSupportVideos: true },
+      },
+      parameters: { subType: 202 },
+    });
+    mockPopiTaskSuccess();
+
+    const ctx = makeCtx(node, {
+      getConnectedInputs: vi.fn().mockReturnValue({
+        images: ["https://static.popi.art/1.png"],
+        videos: ["https://static.popi.art/input.mp4"],
+        audio: [],
+        text: "video prompt",
+        dynamicInputs: {},
+        easeCurve: null,
+      }),
+    });
+    await executeGenerateVideo(ctx);
+
+    const body = JSON.parse(mockFetch.mock.calls[0][1].body);
+    expect(body.parameters.subType).toBe(203);
+  });
+
+  it("normalizes first-last-frame subtype to all-purpose reference when audio input exists", async () => {
+    const node = makeNode({
+      selectedModel: {
+        provider: "popiserver",
+        modelId: "44",
+        displayName: "Seedance",
+        metadata: { isSupportAudios: true },
+      },
+      parameters: { subType: 204 },
+    });
+    mockPopiTaskSuccess();
+
+    const ctx = makeCtx(node, {
+      getConnectedInputs: vi.fn().mockReturnValue({
+        images: ["https://static.popi.art/first.png", "https://static.popi.art/last.png"],
+        videos: [],
+        audio: ["https://static.popi.art/input.mp3"],
+        text: "video prompt",
+        dynamicInputs: {},
+        easeCurve: null,
+      }),
+    });
+    await executeGenerateVideo(ctx);
+
+    const body = JSON.parse(mockFetch.mock.calls[0][1].body);
+    expect(body.parameters.subType).toBe(203);
+  });
+
   it("defaults PopiServer video subtype to image-to-video for exactly one image", async () => {
     const node = makeNode({
       selectedModel: { provider: "popiserver", modelId: "44", displayName: "Seedance" },
       parameters: {},
     });
-    mockFetch.mockResolvedValueOnce({
-      ok: true,
-      json: () => Promise.resolve({ success: true, video: "data:video/mp4;base64,output" }),
-    });
+    mockPopiTaskSuccess();
 
     const ctx = makeCtx(node, {
       getConnectedInputs: vi.fn().mockReturnValue({

src/store/execution/generateVideoExecutor.ts

@@ -54,16 +54,19 @@ function getGeneratedImageUrl(value: unknown): string | undefined {
 
 function normalizeVideoGenerationSubType(
   parameters: Record<string, unknown>,
-  imageCount: number
+  imageCount: number,
+  hasVideoInput: boolean,
+  hasAudioInput: boolean
 ): Record<string, unknown> {
   const subType = Number(parameters.subType);
+  const hasNonImageReference = hasVideoInput || hasAudioInput;
   if (subType === VIDEO_GENERATION_SUB_TYPE_FIRST_LAST_FRAME) {
-    return imageCount === 2
+    return imageCount === 2 && !hasNonImageReference
       ? parameters
       : { ...parameters, subType: VIDEO_GENERATION_SUB_TYPE_ALL_PURPOSE_REFERENCE };
   }
   if (subType === VIDEO_GENERATION_SUB_TYPE_IMAGE_TO_VIDEO) {
-    return imageCount === 1
+    return imageCount === 1 && !hasNonImageReference
       ? parameters
       : { ...parameters, subType: VIDEO_GENERATION_SUB_TYPE_ALL_PURPOSE_REFERENCE };
   }
@@ -72,7 +75,7 @@ function normalizeVideoGenerationSubType(
   }
   return {
     ...parameters,
-    subType: imageCount === 1
+    subType: imageCount === 1 && !hasNonImageReference
       ? VIDEO_GENERATION_SUB_TYPE_IMAGE_TO_VIDEO
       : VIDEO_GENERATION_SUB_TYPE_ALL_PURPOSE_REFERENCE,
   };
@@ -83,6 +86,8 @@ function buildVideoParameters(
   nodeConfig: ReturnType<typeof readVideoGenerationConfig>,
   modelToUse: SelectedModel,
   imageCount: number,
+  hasVideoInput: boolean,
+  hasAudioInput: boolean,
   parametersOverride?: Record<string, unknown>
 ): Record<string, unknown> | undefined {
   let parameters = removeFirstClassGenerationParameters(mergeSchemaDefaults(
@@ -94,7 +99,7 @@ function buildVideoParameters(
   const soundParameterName = getVideoSoundParameterName(modelToUse);
 
   if (modelToUse.provider === "popiserver") {
-    parameters = normalizeVideoGenerationSubType(parameters, imageCount);
+    parameters = normalizeVideoGenerationSubType(parameters, imageCount, hasVideoInput, hasAudioInput);
     const merged = mergeConfiguredParametersOverride(parameters, {
       ...(nodeConfig.aspectRatio
         ? { aspectRatio: nodeConfig.aspectRatio, aspect_ratio: nodeConfig.aspectRatio, ratio: nodeConfig.aspectRatio }
@@ -352,7 +357,15 @@ export async function executeGenerateVideo(
           })
         : { images: preparedInputs.images, videos: [], voices: [] };
       const parameters = sanitizeReferenceSubjectListParameter(
-        buildVideoParameters(nodeData, nodeConfig, modelToUse, supportedInputs.images.length, parametersOverride),
+        buildVideoParameters(
+          nodeData,
+          nodeConfig,
+          modelToUse,
+          supportedInputs.images.length,
+          supportedInputs.videos.length > 0,
+          supportedInputs.voices.length > 0,
+          parametersOverride
+        ),
         nodeData,
         modelToUse
       );