fix(canvas): NewAPI 视频生成 - 拒绝缺少参考视频的纯音频请求

backend/internal/service/provider.go

@@ -1724,6 +1724,10 @@ func newAPIChannel2VideoRequestBody(input canvasGenerationInput) (newAPIVideoReq
 	if len(input.ReferenceImages) > 9 || len(input.ReferenceVideos) > 3 || len(input.ReferenceAudios) > 3 {
 		return newAPIVideoRequest{}, errors.New("NewAPI Video Generations 最多支持 9 张参考图、3 个参考视频和 3 个参考音频")
 	}
+	// NewAPI Video Generations 只接受附着在参考视频上的音频；纯音频请求会被上游拒绝。
+	if len(input.ReferenceAudios) > 0 && len(input.ReferenceVideos) == 0 {
+		return newAPIVideoRequest{}, errors.New("NewAPI Video Generations 的参考音频必须同时提供至少 1 个参考视频；纯音频生视频请切换到支持该模式的渠道")
+	}
 	modelName := strings.ToLower(strings.TrimSpace(input.Config.Model))
 	requiresSingleImage := modelName == "grok-video-1.5" || modelName == "grok-video-1.5-1080p"
 	images := make([]string, 0, len(input.ReferenceImages))

backend/internal/service/provider_test.go

@@ -1064,6 +1064,16 @@ func TestNewAPIChannel2SingleImageModelsRequireOneReference(t *testing.T) {
 	}
 }
 
+func TestNewAPIChannel2RejectsAudioWithoutReferenceVideo(t *testing.T) {
+	_, err := newAPIChannel2VideoRequestBody(canvasGenerationInput{
+		Config:          providerConfig{Model: "grok-image-video", VideoSeconds: "6"},
+		ReferenceAudios: []providerMedia{{ID: "audio-1", URL: "https://example.com/reference.mp3"}},
+	})
+	if err == nil || !strings.Contains(err.Error(), "必须同时提供至少 1 个参考视频") {
+		t.Fatalf("newAPIChannel2VideoRequestBody() error = %v", err)
+	}
+}
+
 func TestNewAPIChannel2SingleImageModelUsesReferenceForStaleTextToVideoMetadata(t *testing.T) {
 	body, err := newAPIChannel2VideoRequestBody(canvasGenerationInput{
 		Config:          providerConfig{Model: "grok-video-1.5", VideoSeconds: "6"},