feat(kling): 接入 Omni 视频兼容接口

2026-07-22 00:25:05 +08:00
parent b04a7d9d3d
commit 5a71643099
10 changed files with 1886 additions and 7 deletions
@@ -338,8 +338,11 @@ func kelingVideoPayload(ctx context.Context, request Request) (map[string]any, s
 	if value, ok := body["cfg_scale"]; ok && numericValue(value, 0) > 0 {
 		payload["cfg_scale"] = value
 	}
-	if boolValue(body, "audio") || boolValue(body, "output_audio") {
-		payload["sound"] = "on"
+	if sound, ok := kelingSoundSetting(body); ok {
+		if sound == "on" && !kelingSupportsGeneratedSound(request.Candidate) {
+			return nil, "", &ClientError{Code: "invalid_parameter", Message: "kling-video-o1 does not support generated audio", StatusCode: http.StatusBadRequest, Retryable: false}
+		}
+		payload["sound"] = sound
 	}
 	if mode := kelingModeByResolution(firstNonEmptyStringValue(body, "resolution", "size")); mode != "" {
 		payload["mode"] = mode
@@ -432,7 +435,7 @@ func (c KelingClient) kelingOmniPayload(ctx context.Context, request Request, to
 		watermarkEnabled = boolValue(watermarkInfo, "enabled")
 	}
 	payload := map[string]any{
-		"model_name":      upstreamModelName(request.Candidate),
+		"model_name":      kelingOmniUpstreamModelName(request.Candidate),
 		"mode":            kelingModeByResolution(firstNonEmptyStringValue(body, "resolution", "size")),
 		"watermark_info":  map[string]any{"enabled": watermarkEnabled},
 		"negative_prompt": strings.TrimSpace(stringFromAny(body["negative_prompt"])),
@@ -458,8 +461,13 @@ func (c KelingClient) kelingOmniPayload(ctx context.Context, request Request, to
 	if voices := mapListFromAny(body["voice_list"]); len(voices) > 0 {
 		payload["voice_list"] = voices
 	}
-	if (boolValue(body, "audio") || boolValue(body, "output_audio")) && !hasVideo {
-		payload["sound"] = "on"
+	if sound, ok := kelingSoundSetting(body); ok {
+		if sound == "on" && !kelingSupportsGeneratedSound(request.Candidate) {
+			return nil, nil, &ClientError{Code: "invalid_parameter", Message: "kling-video-o1 does not support generated audio", StatusCode: http.StatusBadRequest, Retryable: false}
+		}
+		if !hasVideo {
+			payload["sound"] = sound
+		}
 	}
 	if multiShot {
 		payload["multi_shot"] = true
@@ -728,6 +736,18 @@ func kelingIsOmniRequest(request Request) bool {
 		request.Candidate.Capabilities["omni"] != nil
 }

+func kelingOmniUpstreamModelName(candidate store.RuntimeModelCandidate) string {
+	model := strings.TrimSpace(upstreamModelName(candidate))
+	switch strings.ToLower(model) {
+	case "kling-o1":
+		return "kling-video-o1"
+	case "kling-3.0-omni":
+		return "kling-v3-omni"
+	default:
+		return model
+	}
+}
+
 func kelingIs30TurboRequest(request Request) bool {
 	switch strings.ToLower(strings.TrimSpace(upstreamModelName(request.Candidate))) {
 	case "kling-3.0-turbo", "kling-v3-turbo", "kling-3-0-turbo":
@@ -1073,6 +1093,54 @@ func kelingModeByResolution(resolution string) string {
 	}
 }

+func kelingSoundSetting(body map[string]any) (string, bool) {
+	if sound := strings.ToLower(strings.TrimSpace(stringFromAny(body["sound"]))); sound == "on" || sound == "off" {
+		return sound, true
+	}
+	for _, key := range []string{"audio", "output_audio", "generate_audio"} {
+		if enabled, ok := kelingBoolFieldValue(body, key); ok {
+			if enabled {
+				return "on", true
+			}
+			return "off", true
+		}
+	}
+	return "", false
+}
+
+func kelingSupportsGeneratedSound(candidate store.RuntimeModelCandidate) bool {
+	switch strings.ToLower(strings.TrimSpace(upstreamModelName(candidate))) {
+	case "kling-o1", "kling-video-o1":
+		return false
+	default:
+		return true
+	}
+}
+
+func kelingWatermarkEnabled(body map[string]any) bool {
+	if enabled, ok := kelingBoolFieldValue(body, "watermark"); ok {
+		return enabled
+	}
+	info := mapFromAny(body["watermark_info"])
+	if info == nil {
+		return false
+	}
+	enabled, _ := kelingBoolFieldValue(info, "enabled")
+	return enabled
+}
+
+func kelingBoolFieldValue(body map[string]any, key string) (bool, bool) {
+	if body == nil {
+		return false, false
+	}
+	value, ok := body[key]
+	if !ok {
+		return false, false
+	}
+	typed, ok := value.(bool)
+	return typed, ok
+}
+
 func kelingCameraControl(body map[string]any) map[string]any {
 	cameraControl := strings.TrimSpace(stringFromAny(body["camera_control"]))
 	if cameraControl == "" {
@@ -1182,7 +1250,7 @@ func kelingVideoSuccessResult(request Request, upstreamTaskID string, raw map[st
 		if id := strings.TrimSpace(stringFromAny(video["id"])); id != "" {
 			item["id"] = id
 		}
-		if duration := intFromAny(video["duration"]); duration > 0 {
+		if duration := firstPresent(video["duration"]); duration != nil && strings.TrimSpace(stringFromAny(duration)) != "" {
 			item["duration"] = duration
 		}
 		if watermarkURL := strings.TrimSpace(stringFromAny(video["watermark_url"])); watermarkURL != "" {
@@ -1194,11 +1262,15 @@ func kelingVideoSuccessResult(request Request, upstreamTaskID string, raw map[st
 	if created == 0 {
 		created = int(nowUnix())
 	}
+	modelName := upstreamModelName(request.Candidate)
+	if kelingIsOmniRequest(request) {
+		modelName = kelingOmniUpstreamModelName(request.Candidate)
+	}
 	return map[string]any{
 		"id":               upstreamTaskID,
 		"object":           "video.generation",
 		"created":          created,
-		"model":            upstreamModelName(request.Candidate),
+		"model":            modelName,
 		"status":           "succeeded",
 		"upstream_task_id": upstreamTaskID,
 		"data":             items,