kling-v1-6

The exact kling-v1-6 selector identifies the Kling text-to-video profile. Submit JSON to POST /v1/videos/generations with model set to kling-v1-6 and the written scene in prompt. POST /v1/videos also accepts the same generation request. A clear request describes what is visible, what happens during the shot and how the camera observes it. The minimal scene request supplies the model selector and prompt, leaving additional shot controls unspecified.

For a product shot, describe the object, surface, lighting and one visible action, such as a slow turn or a change in camera position. For a scene concept, identify the setting, principal subject and movement through the frame. Keep the action readable within the requested clip. A prompt might describe a small boat crossing a still lake, a wide view from the bank, early morning light and a steady camera following the boat’s direction.

The prompt limit is 500 characters. The optional negative_prompt can contain up to 200 characters and describes unwanted elements. The documented duration choices for this profile are 5 and 10 seconds. The mode values are std and pro. cfg_scale is a number from 0 to 1; the documentation describes higher values as stronger adherence to the text. Choose controls for the exact shot and preserve the submitted values with the result.

The documented camera_control object lists simple, down_back, forward_up, right_turn_forward and left_turn_forward camera types. Its simple configuration names horizontal, vertical, pan, tilt, roll and zoom controls. The minimal text example provides a useful starting point when a scene can be described without a separate camera-control object. Give camera movement a clear purpose in the prompt, such as revealing the subject or following its motion.

Submission starts an asynchronous task. Keep the returned id, X-Ailoot-Task-Id header and polling_url, then use GET on that polling_url with the same Ailoot API key to read the original task. The task state is in data.task_status: submitted and processing are pending, succeed is successful, and failed is unsuccessful. A submission acknowledgement is therefore an intermediate result. On success, retrieve every URL in data.task_result.videos[].url; these are Ailoot-managed result file links. GET /v1/videos/{id}/content retrieves the first output, and its index query parameter selects a returned output by its zero-based index. Keep the status response with the selected clip. If a timeout or result_pending_check response occurs, preserve the original returned identity and check that operation rather than submitting the shot again.

Review the completed footage across its full duration: the intended action, subject appearance, framing and motion all belong to the shot brief. This exact profile provides the text-to-video workflow described here. A returned task that still needs a status check should be followed using its original identity, including when a client request times out; preserve that task information before considering a new generation request.