tusheng video

Aspect ratio (width:height). Upstream constraints: commonly used 16:9 / 9:16 / 1:1, the gateway does not preflight enumeration.

Task completion callback address. POST notification after the upstream task is completed, the public network must be reachable, and the gateway transparently transmits the information as it is.

Lens motion control. If it is not transmitted, the model will intelligently match it; the gateway transparently transmits it and does not pre-check the specific value range.

Generate degrees of freedom (CFG). The larger the value, the stronger the correlation between the generated results and prompt; the gateway does not pre-check the range, and the upstream constraints are usually 0~1.

Video duration (seconds), required. Non-kling-v3: only 5 or 10; kling-v3: 3~15. Supports string or integer, gateway normalizes to string.

The caller's external task ID, used for idempotent association and business reconciliation; non-response body data.task_id.

Enter image URL or Base64, required. As a reference for the first frame of the video; Tusheng video core field. The format/size/resolution is verified by the upstream, the gateway only verifies that it is not empty.

End frame image URL or Base64, optional. Cooperate with image to realize the transition between the first and last frames; if not passed, it will only be generated from the first frame. The gateway transmits transparently and does not pre-check the format.

Generate schema. The billing side defaults to std when not transmitted. 4k only kling-v3; kling-v2-6 requires pro to open sound.

Model name. When not transmitted, gateway accounting defaults to kling-v1, and it is recommended to fill it in explicitly. More kling-v1-5 / kling-v2-1 than text2video. It must be kling-v3 when multi_shot=true; it must be kling-v3 when mode=4k.

Customized storyboard list. Required when shot_type=customize, up to 6 shots. Each mirror prompt≤512, duration≥1; the sum of the durations of each mirror must be equal to the top-level duration (error ≤0.1).

Whether there are multiple lenses. When true, only kling-v3 supports it, and prompt is invalid; when false, shot_type / multi_prompt cannot be passed, and a 400 error will occur if it is passed.

Negative cue words. Describe unwanted content (such as blur, low image quality, deformation, etc.), and the gateway transparently transmits it.

Positive cue words. Required when multi_shot=false (default); invalid when multi_shot=true, 400 will be sent. Describe movement, shots and style, upstream constraint ≤2500 characters.

Storyboarding method. Required when multi_shot=true; cannot be passed when multi_shot=false. customize must be used with multi_prompt; intelligence is automatically removed from the upstream.

Audible switch. The gateway only performs model/schema verification on non-null values; triggers audible billing when on is passed. Only supported by kling-v2-6 / kling-v3; v2-6 requires mode=pro.

Custom sound list. Participate in the billing multiplier (hasVoiceId) when sound=on and is not empty; only image2video path identification, text2video is not included. Requires kling-v2-6 / kling-v3 audio capabilities.