fix(canvas): 短剧分镜 - 支持多镜头并行生成并正确引用角色图声资产

web/test/video-api-contracts.test.ts

@@ -2,6 +2,7 @@ import { describe, expect, test } from "bun:test";
 
 import { defaultConfig } from "../src/stores/use-config-store";
 import { createSeedanceTask } from "../src/services/api/video-provider-seedance";
+import { createMiniMaxVideoTask } from "../src/services/api/video-provider-minimax";
 import type { VideoProviderDeps } from "../src/services/api/video-provider-deps";
 import { unwrapEnvelope, videoResponseTools, videoTaskId } from "../src/services/api/video-response";
 import { isPublicMediaUrl, normalizeVideoResolution, normalizeVideoSeconds, normalizeVideoSize } from "../src/services/api/video-validation";
@@ -100,3 +101,37 @@ describe("Volcengine Ark full-modal references", () => {
         await expect(createSeedanceTask(deps, config as never, model, "跟随节奏", [], [], [{ id: "audio-1", name: "audio.mp3", type: "audio/mpeg", url: "https://cdn.example.com/audio.mp3" }])).rejects.toThrow("不支持纯音频或文本+音频");
     });
 });
+
+describe("MiniMax explicit reference roles", () => {
+    test("project assets use reference_image and reference_audio instead of first frame", async () => {
+        let requestBody: unknown;
+        const deps = {
+            transport: {
+                post: async (_url: string, body: unknown) => {
+                    requestBody = body;
+                    return { task_id: "minimax-task-1" };
+                },
+            },
+            response: videoResponseTools,
+        } as unknown as VideoProviderDeps;
+        const config = { ...defaultConfig, baseUrl: "https://api.minimaxi.com/v1", videoSeconds: "6", vquality: "768P", size: "16:9", videoWatermark: "false" };
+
+        await createMiniMaxVideoTask(
+            deps,
+            config as never,
+            "MiniMax-H3",
+            "保持角色一致",
+            [{ id: "character-1", name: "character.png", type: "image/png", dataUrl: "", url: "https://cdn.example.com/character.png" }],
+            [],
+            [{ id: "voice-1", name: "voice.mp3", type: "audio/mpeg", url: "https://cdn.example.com/voice.mp3" }],
+            { videoEditOperation: "reference_to_video" },
+        );
+
+        expect((requestBody as { content: unknown[] }).content).toEqual([
+            { type: "text", text: "保持角色一致" },
+            { type: "image_url", image_url: { url: "https://cdn.example.com/character.png" }, role: "reference_image" },
+            { type: "audio_url", audio_url: { url: "https://cdn.example.com/voice.mp3" }, role: "reference_audio" },
+        ]);
+        expect((requestBody as { ratio: string }).ratio).toBe("16:9");
+    });
+});

web/test/workflow-shot-generation.test.ts

@@ -0,0 +1,85 @@
+import { expect, test } from "bun:test";
+
+import { defaultConfig } from "../src/stores/use-config-store";
+import { submitBackendGenerationTask, type GenerationTaskDependencies } from "../src/services/api/generation-task";
+import type { GenerationTask } from "../src/services/api/task-center";
+import type { ProjectDetail } from "../src/services/api/projects";
+import { buildShotAssetReferenceContext, resolveShotAssetMentionPrompt } from "../src/pages/projects/detail/workflow-shot-references";
+
+test("shot asset prompt keeps character names and maps character image, voice and scene assets", () => {
+    const detail = {
+        assets: [
+            {
+                id: "character-1",
+                title: "张天昊",
+                category: "character",
+                mediaType: "image",
+                primaryVersionId: "character-version-1",
+                character: {
+                    versionId: "character-version-1",
+                    definition: { voiceLanguage: "普通话", voiceAge: "青年男性", voiceTimbre: "略带疲惫和震惊" },
+                    representations: [{ id: "representation-1", resourceId: "character-image-1", mediaType: "image/png", role: "primary" }],
+                    voice: { profile: { sampleResourceId: "character-audio-1", language: "普通话", timbre: "沉稳" }, instructions: "内心独白语气" },
+                },
+            },
+            {
+                id: "scene-1",
+                title: "坑底场景",
+                category: "environment",
+                mediaType: "image",
+                primaryVersionId: "scene-version-1",
+                storageKey: "resource:scene-image-1",
+            },
+        ],
+        shotReferences: [
+            { shotId: "shot-1", assetVersionId: "character-version-1", status: "linked" },
+            { shotId: "shot-1", assetVersionId: "scene-version-1", status: "linked" },
+        ],
+    } as ProjectDetail;
+
+    const context = buildShotAssetReferenceContext(detail, "shot-1");
+    const prompt = resolveShotAssetMentionPrompt("张天昊在 @[asset:scene-1] 睁开眼睛", context);
+
+    expect(context.referenceImages).toHaveLength(2);
+    expect(context.referenceAudios).toHaveLength(1);
+    expect(context.referenceAudios[0]?.storageKey).toBe("resource:character-audio-1");
+    expect(context.resolvedCharacterVersions).toEqual([{ assetId: "character-1", versionId: "character-version-1" }]);
+    expect(prompt).toContain("张天昊在 图片2 睁开眼睛");
+    expect(prompt).toContain("- 张天昊：人物参考：图片1；声音参考：音频1");
+    expect(prompt).toContain("声音说明：普通话；青年男性；略带疲惫和震惊；内心独白语气");
+    expect(prompt).toContain("- 坑底场景：场景参考：图片2");
+});
+
+test("background generation submission returns after task creation without waiting", async () => {
+    let waitCalls = 0;
+    const task = {
+        id: "task-1",
+        type: "canvas_video",
+        status: "queued",
+        prompt: "角色表演",
+        attempts: 0,
+        createdAt: "2026-08-30T00:00:00.000Z",
+        updatedAt: "2026-08-30T00:00:00.000Z",
+    } satisfies GenerationTask;
+    const dependencies: GenerationTaskDependencies = {
+        createTask: async () => task,
+        waitTask: async () => {
+            waitCalls += 1;
+            throw new Error("should not wait");
+        },
+        runLocal: async () => ({ mode: "video" }),
+        createId: () => "id-1",
+        now: () => "2026-08-30T00:00:00.000Z",
+    };
+
+    const submitted = await submitBackendGenerationTask({
+        projectId: "project-1",
+        mode: "video",
+        prompt: "角色表演",
+        config: { ...defaultConfig, model: "MiniMax-H3", videoModel: "MiniMax-H3" },
+        metadata: { shotId: "shot-1", videoEditOperation: "reference_to_video" },
+    }, dependencies);
+
+    expect(submitted).toBe(task);
+    expect(waitCalls).toBe(0);
+});