A summary of the integration methods, model naming, parameter specifications, and query workflows for domestic video model series.

Domestic Video Models

Domestic video models provide a unified OpenAI-compatible interface, supporting multiple generation modes such as text-to-video, image-to-video, first-and-last-frame video generation, motion control, digital humans, and lip sync.

Quick Start

Video Generation

POST /v1/videos, submit domestic video tasks, supporting multiple generation scenarios.

Task Query

GET /v1/videos/{task_id}, query task status and results.

Model Family

Popular Models

Model Series	Representative Model	Key Capabilities
Jimeng	`jimeng-video-3.0`	Text-to-video, image-to-video, first-and-last-frame video generation
Vidu	`viduq3-pro`	Direct audio-video output, background music, off-peak mode
Kling	`Kling-3.0-Omni`	Motion control, digital humans, lip sync, template effects

Full Model List

Base Models:

Vidu-*, Kling-*, GV-*, OS-*
Hunyuan-*, Mingmou-*, Hailuo-*
SV-*, JV-*, jimeng-video-*

Combined Billing Models:

vidu-q2-pro-reference-1080p-offpeak
kling-3.0-omni-1080p-ref-audio
kling-2.6-motion-pro-1080p
kling-avatar-720p
sv-1.5-pro-1080p-audio

Integration Methods

Domestic video models use the following interfaces in a unified way:

Operation	Endpoint	Description
Create Task	`POST /v1/videos`	Submit a video generation task
Query Status	`GET /v1/videos/{task_id}`	Query task progress and status
Get Content	`GET /v1/videos/{task_id}/content`	Get the video download URL

Note: Different model series may provide multiple integration methods.

Vidu: Official format (/vidu/ent/v2/*), OpenAI format (/v1/videos), unified video (/v1/video/create)

Jimeng: OpenAI format (/v1/videos), unified video (/v1/video/create), Doubao channel

Kling: OpenAI format (/v1/videos)

This page mainly introduces the general specifications of the OpenAI-compatible interface. For model-specific interfaces, please refer to the model documentation below.

Request Parameters

Image Input Description

Different integration methods use different image fields:

Integration Method	Field Name	Format	Description
OpenAI format (multipart)	`input_reference`	file	File upload
OpenAI format (JSON)	`image`	string	Image URL
Unified video	`images`	array	Array of image URLs

Basic Fields

Field	Type	Required	Description
`model`	string	✅	Model name, supports base models or combined billing models
`prompt`	string	✅	Prompt, text description for video generation
`seconds`	string	❌	Video duration, for example `5`, `10`, `15`
`size`	string	❌	Quick size input, for example `720x1280`
`image`	string	❌	Reference image URL (image-to-video scenario)
`images`	array	❌	Reference image array (first-and-last-frame scenario)
`input_reference`	file	❌	Reference file (OpenAI format upload)
`metadata`	object	❌	Extended parameters; it is recommended to pass through upstream native configuration

metadata Extended Parameters

Scenario Type:

Scenario	Value	Description
Motion Control	`motion_control`	Precisely control video motion
Digital Human Generation	`avatar_i2v`	Generate digital human videos
Lip Sync	`lip_sync`	Audio-visual synchronized lip sync
Template Effects	`template_effect`	Apply template effects

Output Configuration (metadata.output_config):

Field	Type	Description
`resolution`	string	Resolution: `720P`, `1080P`
`aspect_ratio`	string	Aspect ratio: `16:9`, `9:16`, `1:1`
`duration`	integer	Duration (seconds)
`audio_generation`	string	Audio generation: `Enabled`, `Disabled`

Other Common Fields:

motion_level: motion level (std/pro)
offpeak: whether off-peak billing is enabled
last_frame_url: the last frame in first-and-last-frame generation
video_url: reference video URL
file_infos: native FileInfos passthrough
ext_info: native ExtInfo string passthrough

Parameter Precedence

Duration Parameters

Top-level seconds
Top-level duration
metadata.seconds / metadata.duration / metadata.video_duration
Default 5

Resolution Parameters

metadata.output_config.resolution
Top-level size
Model default value

Scenario Examples

Text-to-Video

{
  "model": "Kling-3.0-Omni",
  "prompt": "Cyberpunk city night scene, the camera slowly pushes in",
  "seconds": "5",
  "metadata": {
    "output_config": {
      "resolution": "720P",
      "aspect_ratio": "16:9"
    }
  }
}

Image-to-Video

{
  "model": "viduq3-pro",
  "prompt": "Make the character walk forward and smile",
  "image": "https://example.com/character.png",
  "seconds": "5",
  "metadata": {
    "output_config": {
      "resolution": "1080P",
      "aspect_ratio": "9:16"
    }
  }
}

First-and-Last-Frame Video Generation

{
  "model": "Kling-3.0-Omni",
  "prompt": "In a dead-quiet system space, the character is illuminated by a blue panel",
  "seconds": "15",
  "size": "720x1280",
  "metadata": {
    "output_config": {
      "duration": 15,
      "resolution": "720P",
      "aspect_ratio": "9:16",
      "audio_generation": "Enabled"
    },
    "last_frame_url": "https://example.com/last-frame.png"
  }
}

Motion Control

{
  "model": "Kling-3.0-Omni",
  "prompt": "The character waves hello",
  "image": "https://example.com/character.png",
  "seconds": "5",
  "metadata": {
    "scene_type": "motion_control",
    "motion_level": "pro",
    "output_config": {
      "resolution": "1080P",
      "aspect_ratio": "16:9"
    }
  }
}

Model Documentation

Detailed documentation for the following mainstream models is currently provided first:

Jimeng Video

Documentation for generating and querying jimeng-video-3.0 and jimeng-video-2.0.

Vidu Video

The three integration methods for the viduq3-pro, viduq2, and viduq1 series.

Kling Video

Multiple generation modes for Kling-3.0-Omni, Kling-2.6, and Kling-2.5.

Detailed documentation for other models (GV-*, OS-*, Hunyuan-*, Mingmou-*, Hailuo-*, SV-*, JV-*) is being organized. You can refer to Domestic Video Model Generation for general parameter specifications and usage methods.

Authentication

Channel key format:

SubAppId|SecretId|SecretKey
SubAppId|SecretId|SecretKey|Region

Overview of Domestic Video Models