fix(*): MiniMax 视频能力 - 放行参考素材视频生成并统一操作推导

backend/internal/service/model_capability.go

@@ -229,6 +229,7 @@ func DefaultModelCapabilityConfigForModel(protocol string, modelName string) *Mo
 		video.Resolutions = []string{"1080p"}
 		video.DefaultResolution = "1080p"
 	case model.ChannelInterfaceMiniMaxVideo:
+		video.Operations = append(video.Operations, "reference_to_video")
 		video.References.MaxImages = 9
 		video.References.MaxImageBytes = 30 * 1024 * 1024
 		video.References.MaxVideos = 3

backend/internal/service/model_capability_test.go

@@ -83,6 +83,16 @@ func TestDefaultVideoCapabilityUsesProtocolSpecificResolutionTiers(t *testing.T)
 	}
 }
 
+func TestDefaultMiniMaxVideoCapabilitySupportsReferenceGeneration(t *testing.T) {
+	profile := DefaultModelCapabilityConfigForModel("minimax-video", "MiniMax-H3")
+	if profile == nil || profile.Video == nil {
+		t.Fatal("MiniMax video profile = nil")
+	}
+	if !containsCapabilityString(profile.Video.Operations, "reference_to_video") {
+		t.Fatalf("operations = %v, want reference_to_video", profile.Video.Operations)
+	}
+}
+
 func TestCapabilitySpecFromModelCapabilityConfigRestoresLegacyWildcardImageSizes(t *testing.T) {
 	config := &ModelCapabilityConfig{
 		Version: 1,

web/src/lib/model-capabilities.ts

@@ -315,6 +315,7 @@ export function defaultModelCapabilityConfig(protocol?: ModelProtocol, model = "
         video.resolutions = ["768P", "2K"];
         video.defaultResolution = "768P";
         video.watermark = { supported: true, default: false };
+        video.operations.push("reference_to_video");
     }
     return { version: 1, text, image: defaultImageCapabilityConfig(protocol, model), video };
 }

web/src/lib/model-selection.ts

@@ -312,14 +312,13 @@ export function modelGroupReferenceLimits(config: AiConfig, selected: string, ca
 
 export function inferVideoOperation(input: ModelInputSummary) {
     const visualInputCount = input.imageCount + input.characterCount;
-    if (input.audioCount > 0 && visualInputCount === 0 && input.videoCount === 0) return "audio_to_video";
-    if (input.videoCount > 0) return "extend";
+    if (input.audioCount > 0 || input.videoCount > 0 || visualInputCount > 2) return "reference_to_video";
     if (visualInputCount > 0) return "image_to_video";
     return "text_to_video";
 }
 
 export function resolveVideoOperation(input: ModelInputSummary, storedOperation?: string) {
-    if (storedOperation && !["text_to_video", "image_to_video", "audio_to_video", "extend"].includes(storedOperation)) return storedOperation;
+    if (storedOperation && !["text_to_video", "image_to_video", "audio_to_video", "extend", "reference_to_video"].includes(storedOperation)) return storedOperation;
     return inferVideoOperation(input);
 }
 
@@ -327,6 +326,7 @@ function videoOperationLabel(operation: string) {
     if (operation === "text_to_video") return "文生视频";
     if (operation === "image_to_video") return "图生视频";
     if (operation === "audio_to_video") return "音频生视频";
+    if (operation === "reference_to_video") return "参考素材生视频";
     if (operation === "extend") return "视频续写";
     return "当前生成模式";
 }

web/src/services/api/generation-task.ts

@@ -6,6 +6,7 @@ import { LOCAL_DREAMINA_WAIT_STOPPED_CODE, LocalDreaminaGenerationClientError, r
 import { isLocalDreaminaBackgroundTask, localDreaminaTaskId, projectLocalDreaminaTask, stripLocalDreaminaTaskPrefix } from "@/services/local-dreamina-task-projection";
 import { modelCapabilityConfigFor } from "@/lib/model-capabilities";
 import { grokImagePromptLimitError } from "@/lib/grok-image-prompt-limit";
+import { resolveVideoOperation } from "@/lib/model-selection";
 import { modelOptionName, resolveModelChannel, resolveModelRequestConfig, type AiConfig } from "@/stores/use-config-store";
 import { useLocalDreaminaModelStore } from "@/stores/use-local-dreamina-model-store";
 import type { ReferenceImage } from "@/types/image";
@@ -268,10 +269,13 @@ async function runLocalDreaminaGeneration(options: BackendGenerationTaskOptions,
 
 function generationOperation(options: BackendGenerationTaskOptions) {
     if (options.mode !== "video") return options.mode;
-    const imageCount = options.referenceImages?.length ?? 0;
-    if ((options.referenceVideos?.length ?? 0) > 0 || (options.referenceAudios?.length ?? 0) > 0 || imageCount > 2) return "reference_to_video";
-    if (imageCount > 0) return "image_to_video";
-    return "text_to_video";
+    return resolveVideoOperation({
+        textCount: 0,
+        imageCount: options.referenceImages?.length ?? 0,
+        videoCount: options.referenceVideos?.length ?? 0,
+        audioCount: options.referenceAudios?.length ?? 0,
+        characterCount: 0,
+    }, options.metadata?.videoEditOperation as string | undefined);
 }
 
 export function isGenerationTaskCancelled(error: unknown, signal?: AbortSignal) {