Generic Video Parameter Guide

Create task: POST /v1/videos

Equivalent create routes:

POST /v1/video/generations

POST /v1/videos/generations

Query task: GET /v1/videos/{task_id}

Equivalent query routes:

GET /v1/video/generations/{task_id}

GET /v1/videos/generations/{task_id}

1. Supported Models

This document applies to the following video models, for example:

Kling-3.0-1080p

Kling-3.0-omni-1080p

Vidu-q2-720p

Vidu-q2-turbo-720p

Hailuo-2.3-fast-768p

GV-3.1-1080p

OS-2.0-720p

2. General Request Body

{
  "model": "Kling-3.0-omni-1080p",
  "seconds": "5",
  "prompt": "A girl dances while holding a puppy",
  "images": [
    "https://ark-project.tos-cn-beijing.volces.com/doc_image/seepro_first_frame.jpeg",
    "https://ark-project.tos-cn-beijing.volces.com/doc_image/seelite_ref_2.png"
  ],
  "metadata": {
    "aspect_ratio": "16:9",
    "negative_prompt": "low quality, blur, shake, watermark, subtitles, deformed fingers",
    "enhance_prompt": "Enabled",
    "input_region": "Mainland",
    "output_config": {
      "AudioGeneration": "Enabled",
      "PersonGeneration": "AllowAdult",
      "InputComplianceCheck": "Disabled",
      "OutputComplianceCheck": "Disabled"
    }
  }
}

3. Top-Level Fields

3.1 model

Required.

Use model name + version + resolution suffix.

Examples:

Kling-2.5-720p

Kling-3.0-omni-1080p

Vidu-q2-720p

Vidu-q2-turbo-720p

Hailuo-2.3-fast-768p

GV-3.1-1080p

OS-2.0-720p

Notes:

Resolution is controlled by model.

metadata.output_config.Resolution is not allowed to override it.

3.2 seconds

seconds is required for all video models.

Kling 2.5: 5 or 10

Kling O1: 3-10

Kling 3.0 / 3.0-Omni: 3-15

Hailuo 2.3 / 2.3-fast

768P: 6 or 10

1080P: 10

Vidu Q2 series: 1-10

Vidu Q3 series: 1-16

GV: 8

OS: 4, 8, 12

Common errors:

seconds is required

Kling 2.5 seconds must be 5 or 10

GV seconds must be 8

3.3 prompt

Text prompt.

Rules:

prompt is required when both images and videos are empty.

For Kling 3.0 / 3.0-Omni, when multi_shot=true, the top-level prompt is overridden by multi_prompt shot prompts.

3.4 images

Use the images array for reference images.

Rules:

Only http:// or https:// URLs are supported.

The single-field image input is not supported.

Base64 is not supported.

Common errors:

image is not supported; use images (URL only)

images must be http(s) URLs

images rules by model

Vidu

q2: up to 7 reference images

q2-pro / q2-turbo: up to 2 images

images[0] is used as the first frame

images[1] is used as the last frame

Hunyuan / Mingmou

No reference images are supported.

Hailuo

2.3 / 2.3-fast: up to 1 reference image

Kling

2.5 + 720P: up to 1 reference image

2.5 + 1080P: up to 2 images

O1 / 3.0 / 3.0-Omni: supports more advanced reference input, see the advanced reference rules below

GV

3.1 / 3.1-fast: up to 2 images

OS

2.0: up to 1 reference image

3.5 videos

Use the videos array for reference videos.

Rules:

Only http:// or https:// URLs are supported.

The single-field video input is not supported.

Currently supported only for:

Kling O1

Kling 3.0

Kling 3.0-Omni

Common errors:

video is not supported; use videos (URL only)

videos is only supported for Kling O1 / 3.0 / 3.0-Omni

4. metadata

The following metadata fields are currently supported:

negative_prompt

enhance_prompt

input_region

aspect_ratio

scene_type

ext_info

SceneType

ExtInfo

output_config

vidu_q2_object_ids

vidu_q2_voice_map

Notes:

Any other non-empty metadata field will be rejected:

metadata.<k> is not supported

4.1 negative_prompt

Negative prompt.

4.2 enhance_prompt

Whether to automatically enhance the prompt.

Allowed values:

Enabled

Disabled

4.3 input_region

Region of the input material URL.

Allowed values:

Mainland

Oversea

Common errors:

metadata.input_region must be Oversea or Mainland

4.4 aspect_ratio

Video aspect ratio.

Kling

Supported for text-to-video:

16:9

9:16

1:1

Vidu

q2 supports:

16:9

9:16

4:3

3:4

1:1

q2-pro / q2-turbo / q3* support:

16:9

9:16

1:1

GV

16:9

9:16

OS

Supported only for text-to-video:

16:9

9:16

Hailuo

Not supported

Common errors:

metadata.aspect_ratio is invalid

Hailuo does not support aspect_ratio

GV aspect_ratio must be 16:9 or 9:16

OS aspect_ratio is only supported for text-to-video

4.5 scene_type / ext_info

Used to pass workflow scenes or advanced additional parameters.

Examples:

scene_type=motion_control

scene_type=template_effect

ext_info.AdditionalParameters.multi_shot

ext_info.AdditionalParameters.element_list

Behavior notes:

metadata.scene_type and metadata.SceneType are both supported

metadata.ext_info / metadata.ExtInfo support string or object input

If AdditionalParameters is an object, the system automatically converts it to the required JSON string

5. Vidu:q2 Multi-Subject Parameters

Supported only by Vidu:q2:

metadata.vidu_q2_object_ids

metadata.vidu_q2_voice_map

5.1 vidu_q2_object_ids

Used to assign a subject ID to each reference image. You can reference the subject later in prompt using @subjectId.

Rules:

Must be an array

Must cover the actual number of reference images being used

Each item must be a non-empty string

Common errors:

metadata.vidu_q2_object_ids must be an array

metadata.vidu_q2_object_ids length must match images (at least first 1-7) for Vidu:q2

5.2 vidu_q2_voice_map

Used to assign a voice to a subject.

Rules:

Must be an object

Must be used together with vidu_q2_object_ids

Each value must be a non-empty string

Common errors:

metadata.vidu_q2_voice_map must be an object

metadata.vidu_q2_voice_map requires metadata.vidu_q2_object_ids

6. metadata.output_config

Currently supported fields:

OffPeak

AudioGeneration

PersonGeneration

InputComplianceCheck

OutputComplianceCheck

LogoAdd

EnhanceSwitch

6.1 OffPeak

Supported only by Vidu.

Allowed values:

Enabled

Disabled

Common errors:

output_config.OffPeak must be Enabled or Disabled

output_config.OffPeak is only supported for modelName Vidu

6.2 AudioGeneration

Allowed values:

Enabled

Disabled

Supported models:

GV

OS

Vidu

Seedance

6.3 PersonGeneration

Allowed values:

AllowAdult

Disallowed

6.4 InputComplianceCheck / OutputComplianceCheck

Allowed values:

Enabled

Disabled

6.5 LogoAdd

Supported only by Vidu.

Allowed values:

Enabled

Disabled

Compatible with: 0/1

Compatible with: true/false

Common errors:

output_config.LogoAdd must be Enabled or Disabled

output_config.LogoAdd is only supported for modelName Vidu

7. Advanced Kling O1 / 3.0 / 3.0-Omni Reference Rules

When using the following models:

Kling O1

Kling 3.0

Kling 3.0-Omni

More advanced reference input is supported, with strict validation.

7.1 Image / Video / Element Limits

When reference videos are provided

Reference images must not exceed 4

images + element_list must not exceed 4

When reference videos are not provided

Reference images must not exceed 7

images + element_list must not exceed 7

First-frame or first-last-frame mode

element_list supports at most 3 elements

7.2 multi_shot / multi_prompt Rules

When AdditionalParameters.multi_shot=true:

shot_type must be customize

multi_prompt must contain 1-6 items

Each multi_prompt.prompt must not exceed 512 characters

Each multi_prompt.index must be greater than or equal to 1

Each multi_prompt.duration must be greater than or equal to 1

Each multi_prompt.duration must not exceed the total task duration

The sum of all multi_prompt.duration values must equal seconds

When multi_shot is not enabled:

shot_type must not be provided

Non-empty multi_prompt must not be provided

Parameter Description

Generic Video Parameter Guide#

1. Supported Models#

2. General Request Body#

3. Top-Level Fields#

3.1 model#

3.2 seconds#

3.3 prompt#

3.4 images#

images rules by model#

Vidu#

Hunyuan / Mingmou#

Hailuo#

Kling#

GV#

OS#

3.5 videos#

4. metadata#

4.1 negative_prompt#

4.2 enhance_prompt#

4.3 input_region#

4.4 aspect_ratio#

Kling#

Vidu#

GV#

OS#

Hailuo#

4.5 scene_type / ext_info#

5. Vidu:q2 Multi-Subject Parameters#

5.1 vidu_q2_object_ids#

5.2 vidu_q2_voice_map#

6. metadata.output_config#

6.1 OffPeak#

6.2 AudioGeneration#

6.3 PersonGeneration#

6.4 InputComplianceCheck / OutputComplianceCheck#

6.5 LogoAdd#

7. Advanced Kling O1 / 3.0 / 3.0-Omni Reference Rules#

7.1 Image / Video / Element Limits#

When reference videos are provided#

When reference videos are not provided#

First-frame or first-last-frame mode#

7.2 multi_shot / multi_prompt Rules#

Generic Video Parameter Guide

1. Supported Models

2. General Request Body

3. Top-Level Fields

3.1 model

3.2 seconds

3.3 prompt

3.4 images

images rules by model

Vidu

Hunyuan / Mingmou

Hailuo

Kling

GV

OS

3.5 videos

4. metadata

4.1 negative_prompt

4.2 enhance_prompt

4.3 input_region

4.4 aspect_ratio

Kling

Vidu

GV

OS

Hailuo

4.5 scene_type / ext_info

5. Vidu:q2 Multi-Subject Parameters

5.1 vidu_q2_object_ids

5.2 vidu_q2_voice_map

6. metadata.output_config

6.1 OffPeak

6.2 AudioGeneration

6.3 PersonGeneration

6.4 InputComplianceCheck / OutputComplianceCheck

6.5 LogoAdd

7. Advanced Kling O1 / 3.0 / 3.0-Omni Reference Rules

7.1 Image / Video / Element Limits

When reference videos are provided

When reference videos are not provided

First-frame or first-last-frame mode

7.2 multi_shot / multi_prompt Rules