Core capabilities
- Text to video — Generate a 3–10 second video from a text prompt
- Image and element reference — Preserve characters, objects, scenes, and style with images or Legacy element IDs
- Video editing — Add, remove, or modify elements and change composition, viewpoint, style, color, or weather
- Video reference — Reference content, motion, style, or camera movement to create a previous or next shot
- First/end frames — Guide a transition with a first frame or with both first and end frames
Use
model_name: "kling-o1". Kling O1 supports 3–10 seconds, 720p (std), and 1080p (pro). It does not support multi-shot generation or 4K.Quick example
cURL
Multimodal prompts
Use<<<image_1>>>, <<<element_1>>>, and <<<video_1>>> in prompt to reference items in the corresponding arrays. Numbering starts at 1 and follows array order.
Video editing and reference
refer_type: "base"edits the input video. It cannot be combined with first/end frames, anddurationis ignored because output duration follows the input.refer_type: "feature"references content, action, style, or camera movement and can generate a previous or next shot.- Set
keep_original_soundtoyesorno. - Set
soundtooffwhenevervideo_listis present.
Media limits
- Images accept URL or raw Base64 without a data URI prefix. Use JPG/JPEG/PNG up to 10 MB, at least 300 px per side, and an aspect ratio from 1:2.5 to 2.5:1.
- An end frame requires a first frame. Kling O1 does not support first/end-frame mode when
image_listcontains more than two images. - First/end-frame mode cannot use elements.
- Videos must use an accessible MP4/MOV URL, at least 3 seconds, dimensions 720–2160 px, 24–60 fps, at most one file up to 200 MB, and aspect ratio 1:2.5 to 2.5:1.
- Without a reference video and with only multi-image elements, images plus elements cannot exceed 7; with a reference video, the total cannot exceed 4.
API Reference
View the complete Legacy schema, scenario examples, and task query responses.