fix: align video generation payloads

2026-05-14 00:14:54 +08:00
parent f254551522
commit 3225833f96
11 changed files with 702 additions and 188 deletions
@@ -329,6 +329,12 @@ func TestVolcesClientVideoSubmitsAndPollsTask(t *testing.T) {
 	var gotModel string
 	var gotText string
 	var gotFirstFrameRole string
+	var gotDuration float64
+	var gotRatio string
+	var gotResolution string
+	var gotSeed float64
+	var gotCameraFixed bool
+	var gotWatermark bool
 	var submittedRemoteTaskID string
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		gotAuth = r.Header.Get("Authorization")
@@ -343,6 +349,17 @@ func TestVolcesClientVideoSubmitsAndPollsTask(t *testing.T) {
 			if body["prompt"] != nil || body["first_frame"] != nil {
 				t.Fatalf("video convenience fields leaked upstream: %+v", body)
 			}
+			for _, key := range []string{"duration_seconds", "aspect_ratio", "audio", "cameraFixed"} {
+				if _, ok := body[key]; ok {
+					t.Fatalf("volces video task body should not include top-level %s: %+v", key, body)
+				}
+			}
+			gotDuration, _ = body["duration"].(float64)
+			gotRatio, _ = body["ratio"].(string)
+			gotResolution, _ = body["resolution"].(string)
+			gotSeed, _ = body["seed"].(float64)
+			gotCameraFixed, _ = body["camera_fixed"].(bool)
+			gotWatermark, _ = body["watermark"].(bool)
 			content, _ := body["content"].([]any)
 			textItem, _ := content[0].(map[string]any)
 			gotText, _ = textItem["text"].(string)
@@ -375,6 +392,10 @@ func TestVolcesClientVideoSubmitsAndPollsTask(t *testing.T) {
 			"first_frame":  "https://example.com/first.png",
 			"duration":     6,
 			"aspect_ratio": "16:9",
+			"resolution":   "720p",
+			"seed":         11,
+			"cameraFixed":  false,
+			"watermark":    true,
 		},
 		Candidate: store.RuntimeModelCandidate{
 			BaseURL:           server.URL,
@@ -406,10 +427,11 @@ func TestVolcesClientVideoSubmitsAndPollsTask(t *testing.T) {
 	if gotModel != "doubao-seedance-2-0-260128" || gotFirstFrameRole != "first_frame" {
 		t.Fatalf("unexpected submitted model=%s role=%s", gotModel, gotFirstFrameRole)
 	}
-	for _, fragment := range []string{"A clean product reveal", "--dur 6", "--ratio 16:9", "--watermark false", "--seed -1"} {
-		if !strings.Contains(gotText, fragment) {
-			t.Fatalf("expected text to contain %q, got %q", fragment, gotText)
-		}
+	if gotText != "A clean product reveal" {
+		t.Fatalf("video params should not be appended to prompt text, got %q", gotText)
+	}
+	if gotDuration != 6 || gotRatio != "16:9" || gotResolution != "720p" || gotSeed != 11 || gotCameraFixed != false || gotWatermark != true {
+		t.Fatalf("unexpected submitted video params duration=%v ratio=%s resolution=%s seed=%v camera_fixed=%v watermark=%v", gotDuration, gotRatio, gotResolution, gotSeed, gotCameraFixed, gotWatermark)
 	}
 	data, _ := response.Result["data"].([]any)
 	item, _ := data[0].(map[string]any)
@@ -418,6 +440,147 @@ func TestVolcesClientVideoSubmitsAndPollsTask(t *testing.T) {
 	}
 }

+func TestVolcesVideoBodyAllowsOnlyTaskPayloadFields(t *testing.T) {
+	body := volcesVideoBody(Request{
+		Kind:      "videos.generations",
+		ModelType: "omni_video",
+		Model:     "豆包Seedance",
+		Body: map[string]any{
+			"model":                 "豆包Seedance",
+			"duration":              8,
+			"duration_seconds":      8,
+			"aspect_ratio":          "9:16",
+			"resolution":            "720p",
+			"audio":                 true,
+			"callback_url":          "https://example.com/callback",
+			"returnLastFrame":       true,
+			"executionExpiresAfter": 3600,
+			"draft":                 false,
+			"cameraFixed":           false,
+			"watermark":             true,
+			"seed":                  -1,
+			"task_id":               "local-task-id",
+			"runMode":               "simulation",
+			"fps":                   24,
+			"content": []any{
+				map[string]any{"type": "text", "text": "Use <<<element_1>>> in a product reveal"},
+				map[string]any{
+					"type": "element",
+					"element": map[string]any{
+						"inline_element": map[string]any{
+							"name":              "subject",
+							"frontal_image_url": "https://example.com/subject.png",
+							"refer_images":      []any{map[string]any{"url": "https://example.com/side.png", "slot_key": "side"}},
+						},
+					},
+				},
+				map[string]any{
+					"type":      "image_url",
+					"role":      "unexpected_role",
+					"name":      "drop-me",
+					"image_url": map[string]any{"url": "https://example.com/ref.png", "extra": "drop-me"},
+				},
+				map[string]any{
+					"type":     "video_url",
+					"duration": 3,
+					"video_url": map[string]any{
+						"url":                 "https://example.com/ref.mp4",
+						"refer_type":          "feature",
+						"keep_original_sound": "yes",
+						"extra":               "drop-me",
+					},
+				},
+				map[string]any{
+					"type":      "audio_url",
+					"audio_url": map[string]any{"url": "https://example.com/ref.mp3", "extra": "drop-me"},
+				},
+			},
+		},
+		Candidate: store.RuntimeModelCandidate{
+			ModelName:         "豆包Seedance",
+			ProviderModelName: "doubao-seedance-2-0-260128",
+			Credentials:       map[string]any{"apiKey": "volces-key"},
+		},
+	})
+
+	allowedTopLevel := map[string]bool{
+		"model": true, "content": true, "callback_url": true, "return_last_frame": true, "execution_expires_after": true,
+		"generate_audio": true, "draft": true, "resolution": true, "ratio": true, "duration": true,
+		"seed": true, "camera_fixed": true, "watermark": true,
+	}
+	for key := range body {
+		if !allowedTopLevel[key] {
+			t.Fatalf("unexpected top-level volces field %q in %+v", key, body)
+		}
+	}
+	if body["model"] != "doubao-seedance-2-0-260128" ||
+		body["generate_audio"] != true ||
+		body["callback_url"] != "https://example.com/callback" ||
+		body["return_last_frame"] != true ||
+		body["execution_expires_after"] != 3600 ||
+		body["draft"] != false ||
+		body["resolution"] != "720p" ||
+		body["ratio"] != "9:16" ||
+		body["duration"] != 8 ||
+		body["seed"] != -1 ||
+		body["camera_fixed"] != false ||
+		body["watermark"] != true {
+		t.Fatalf("unexpected direct video fields: %+v", body)
+	}
+
+	content, ok := body["content"].([]map[string]any)
+	if !ok || len(content) != 5 {
+		t.Fatalf("unexpected sanitized content: %#v", body["content"])
+	}
+	text := content[0]
+	if text["type"] != "text" || strings.Contains(text["text"].(string), "--dur") || strings.Contains(text["text"].(string), "--ratio") {
+		t.Fatalf("video params should not be appended to the text item: %+v", text)
+	}
+	elementImage := content[1]
+	if elementImage["type"] != "image_url" || elementImage["role"] != "reference_image" {
+		t.Fatalf("referenced element should be converted to reference image: %+v", elementImage)
+	}
+	imageURL, _ := elementImage["image_url"].(map[string]any)
+	if imageURL["url"] != "https://example.com/subject.png" || len(imageURL) != 1 {
+		t.Fatalf("element image payload should only include url: %+v", imageURL)
+	}
+	referenceImage := content[2]
+	if referenceImage["role"] != "reference_image" || referenceImage["name"] != nil {
+		t.Fatalf("image references should be role-normalized and scrubbed: %+v", referenceImage)
+	}
+	videoItem := content[3]
+	videoURL, _ := videoItem["video_url"].(map[string]any)
+	if videoItem["role"] != "reference_video" || videoURL["url"] != "https://example.com/ref.mp4" || videoURL["refer_type"] != "feature" || videoURL["extra"] != nil {
+		t.Fatalf("video references should keep only allowed nested fields: %+v", videoItem)
+	}
+	audioItem := content[4]
+	audioURL, _ := audioItem["audio_url"].(map[string]any)
+	if audioItem["role"] != "reference_audio" || audioURL["url"] != "https://example.com/ref.mp3" || len(audioURL) != 1 {
+		t.Fatalf("audio references should keep only url: %+v", audioItem)
+	}
+}
+
+func TestVolcesVideoBodyPrefersFramesOverDuration(t *testing.T) {
+	body := volcesVideoBody(Request{
+		Kind:      "videos.generations",
+		ModelType: "video_generate",
+		Body: map[string]any{
+			"prompt":   "A quick camera move",
+			"duration": 8,
+			"frames":   57,
+		},
+		Candidate: store.RuntimeModelCandidate{
+			ProviderModelName: "doubao-seedance-1-0-pro-250528",
+		},
+	})
+	if body["frames"] != 57 {
+		t.Fatalf("frames should be passed through as the official duration control: %+v", body)
+	}
+	if _, ok := body["duration"]; ok {
+		t.Fatalf("duration should not be sent when frames is present: %+v", body)
+	}
+}
+
 func TestVolcesClientVideoResumePollsExistingTaskID(t *testing.T) {
 	var submitCalled bool
 	var pollPath string