Generate Video

POST /v1/platform/video

Submit an AI video generation task. Supports text-to-video, image-to-video, video continuation and more. Tasks execute asynchronously, returns task_id. Poll for results via the Query Task API

⚠️ File URLs generated by this API are only retained for 7 days. Please download and save them promptly.

⚠️ The video API parameters are highly similar to the image API. The main difference is: model must start with video-; each item in inputs adds category, reference_type and other video-specific fields; output adds duration, audio_generation and other fields.

Request Parameters

Parameter	Type	Required	Description
model	string	Required	Model identifier, must start with video-. Format: video-{model_name}-{version}, e.g. video-kling-2.1.
prompt	string	Required	Text description for video generation. Describe scene content, actions, style, etc.
negative_prompt	string	Optional	Negative prompt describing content you don't want in the video.
enhance_prompt	string	Optional	Whether to enable automatic prompt enhancement. 可选值：enabled / disabled
generation_mode	string	Optional	Generation mode, e.g. text-to-video, image-to-video. Available values depend on the model.
inputs	array	Optional	List of input files. Can include reference images, reference videos, or tail frame images. ↳ inputs[i] child fields
subject_inputs	array	Optional	List of subject reference images for consistent character generation (supported by Kling / Vidu). ↳ subject_inputs[i] child fields
output	object	Optional	Output configuration controlling video duration, audio, resolution, etc. ↳ output child fields (includes all image API fields; below are video-specific additions)
scene_type	string	Optional	Scene type for generation optimization in specific scenarios: • Kling：motion_control（motion control）、avatar_i2v（digital human）、lip_sync（lip sync） • Vidu：template_effect（special effects template）
input_region	string	Optional	Source region of input material. 可选值：mainland / oversea
session_id	string	Optional	Deduplication identifier. If a request with the same session_id was made within 3 days, this request returns an error. Max 50 characters. Leave empty to skip deduplication.
session_context	string	Optional	Source context for passing user request info. The field value is returned in task callbacks. Max 1000 characters.
ext_info	object	Optional	Reserved field for special purposes. The specific format is determined by the model used.

inputs[i] child fields

Field Name	Type	Required	Description
url	string	Required	Publicly accessible URL of the input material. Recommend images under 10 MB. Supported formats: jpeg, jpg, png.
category	string	Optional	File category. 可选值：image / video
reference_type	string	Optional	Reference type. GV model: asset (material reference), style (style reference). Kling model when category is video: feature (feature reference video), base (video to be edited).
object_id	string	Optional	Subject ID, applicable to Vidu-q2 model. Fill in when identifying a subject in the image; can be referenced via @subjectID. Valid when category is image.
voice_id	string	Optional	Voice ID, applicable to Vidu-q2 model. When all images carry a subject ID, you can assign a voice to each subject. Valid when category is image.
keep_original_sound	string	Optional	Whether to preserve the original audio track of the input video. Valid when category is video. 可选值：enabled / disabled
usage	string	Optional	Purpose of the input file, used to distinguish first-frame, last-frame, or reference generation. Default: reference. 可选值：firstFrame / reference / lastFrame firstFrame: first frame for first/last-frame video or image-to-video; reference: reference-guided video; lastFrame: last frame for first/last-frame video.
text	string	Optional	Subject name, only valid for PixVerse multi-image (subject) reference mode. Referenced in the Prompt via @Text, e.g. @cat running.

subject_inputs[i] child fields

Field Name	Type	Required	Description
url	string	Required	Publicly accessible URL of the subject reference image.
id	string	Optional	Fixed subject ID. Kling: Required; Vidu: Optional.
name	string	Optional	Fixed subject name. Vidu: Required; Kling: Optional.

output child fields (includes all image API fields; below are video-specific additions)

Field Name	Type	Required	Description
duration	integer	Optional	Duration of the generated video in seconds. Available values per model: • Kling：5、10，Default 5 • Hailuo：6、10，Default 6 • Vidu：1~10 • GV：8，Default 8 • OS：4、8、12，Default 8
resolution	string	Optional	Resolution of the generated video. Available values per model: • Kling：720P、1080P，Default 720P • Hailuo：768P、1080P，Default 768P • Vidu：720P、1080P，Default 720P • GV：720P、1080P，Default 720P • OS：720P
aspect_ratio	string	Optional	Specify the aspect ratio of the generated video. Available values per model: • Kling (text-to-video): 16:9、9:16、1:1，Default 16:9 • Vidu (text-to-video / reference image): 16:9、9:16、4:3、3:4、1:1 (q2 only supports 4:3, 3:4) • GV：16:9、9:16，Default 16:9 • OS (text-to-video): 16:9、9:16，Default 16:9 • Hailuo：Not supported yet
audio_generation	string	Optional	Whether to generate audio. Supported models:GV、OS、Vidu。Default: disabled. Equivalent to the boolean audio field; this field takes priority. 可选值：enabled / disabled
audio	boolean	Optional	audio_generation Shorthand for audio_generation. Setting true is equivalent to "enabled". Ignored if audio_generation is already set.
enhance_switch	string	Optional	Whether to enable video enhancement. When the selected resolution exceeds the model's native capability, enhancement is enabled by default. You may output at lower resolution and apply enhancement to achieve higher resolution. 可选值：enabled / disabled
frame_interpolate	string	Optional	Whether to enable Vidu smart frame interpolation for smoother video. Currently only supports the Vidu model. 可选值：enabled / disabled
media_name	string	Optional	Output media filename, max 64 characters. Defaults to a system-generated filename.
class_id	integer	Optional	Category ID for media classification management. Default: 0 (other).

Request Example

JSON (Text-to-Video)

{
  "model":  "video-jimeng-3.0pro",
  "prompt": "A cat running on the beach, golden sunset, slow motion",
  "enhance_prompt": "enabled",
  "output": {
    "duration":          5,
    "aspect_ratio":     "16:9",
    "audio_generation": "enabled",
    "frame_interpolate":"enabled",

  }
}

JSON (Image-to-Video with head/tail frames)

{
  "model":  "video-kling-2.1",
  "prompt": "Camera slowly zooms in, flowers gently swaying in the breeze",
  "inputs": [
    {
      "url":      "https://example.com/first-frame.jpg",
      "category":"image"
    },
    {
      "url":      "https://example.com/last-frame.jpg",
      "category":"last_frame"
    }
  ],
  "output": {
    "duration":     5,
    "aspect_ratio": "16:9"
  }
}

Response Parameters

Field Name	Type	Description
task_id	string	Unique task ID for subsequent task status queries.
status	string	Initial task status, usually queued.