Supported Models and ParametersVideo Models
Model Overview
Review shared video-model fields, differences, and entry pages for each video model.
Current public video models also use POST /v1/generations. The main differences are reference-image fields, resolution enums, and default duration.
The fields most likely to cause integration mistakes are reference_videos, reference_images, resolution, and duration.
Shared Video Fields
| Field | Type | Required | Description |
|---|---|---|---|
model | string | Yes | Target video model ID |
prompt | string | Yes | Video prompt |
reference_videos | string[] | No | Input video URL list, up to 3 files; enables video-input pricing |
input_video_duration | number | No | Processed input seconds; defaults to output duration |
reference_audios | string[] | No | Reference audio URL list, up to 3 files |
generate_audio | boolean | No | Whether to generate audio; defaults to true |
aspect_ratio | string | No | Output aspect ratio |
resolution | string | No | Output resolution |
duration | integer | No | Duration in seconds |
Video Model Differences
| Model | Region | Best For | Reference Inputs | resolution | Defaults | Details |
|---|---|---|---|---|---|---|
seedance-2 | Global / CN | Standard and high-resolution video | Image / video input | 480p 720p 1080p 4K | aspect_ratio=16:9 resolution=720p duration=5 | View |
seedance-2-fast | Global / CN | Speed-first generation | Image / video input | 480p 720p | aspect_ratio=16:9 resolution=720p duration=5 | View |
seedance-2-mini | Global / CN | Lightweight generation | Image / video input | 480p 720p | aspect_ratio=adaptive resolution=480p duration=4 | View |
seedance-2-5 | Global / CN | Seedance 2.5 | Reference images / first and last frames | 480p 720p 1080p | resolution=720p duration=5 | View |
kling-3.0-turbo | Global / CN | Speed-first | Reference images | 720p 1080p | resolution=720p duration=5 | View |
kling-v2-6 | Global / CN | Standard/pro tiers | Reference images | Via mode | mode=std duration=5 | View |
kling-v3 | Global / CN | High quality and 4K | Reference images | Via mode | mode=std duration=5 | View |
kling-v3-omni | Global / CN | Multimodal video | Image / video URLs | Via mode | mode=std duration=5 | View |
kling-video-o1 | Global / CN | Omni video | Image / video URLs | Via mode | mode=std duration=5 | View |
gemini_omni_flash | Global / CN | Gemini Omni | Reference images | 720p 1080p | duration=10 | View |
How to Choose a Video Model
- Choose
seedance-2for higher resolutions,seedance-2-fastfor speed, andseedance-2-minifor lightweight jobs. - Kling quality tiers use
mode:std,pro, and4kwhere declared by the model schema. - Kling V3 Omni and Kling Video O1 accept
reference_videos; includeinput_video_durationwith video input. - Available models and parameters vary by region, API-key access, and route health. Treat
GET /v1/modelsas authoritative.
Smallest Working Video Example
{
"model": "seedance-2",
"prompt": "A product teaser video, clean white background, smooth motion, premium style"
}