mingyan

Omni Flash - Create Task

Create Omni Flash video generation tasks with text, first-frame, first/last-frame, multi-image reference, and video reference inputs

POST/kyyReactApiServer/v1/omni-flash/videos

Omni Flash Create Video Task

Create a video generation task with omni-flash. The public API uses one model name, and the system automatically selects the appropriate upstream mode based on whether input_video is provided.

Authentication

Get Key

All requests require a Bearer token in the request header:

cURL

Authorization: Bearer {{key}}

Base URL

https://zcbservice.aizfw.cn/kyyReactApiServer

baseUrl is the shared prefix for all public APIs. The api field in the current page frontmatter shows the full endpoint. Use this baseUrl as the common prefix when reading or composing request paths.

Request Parameters

modelbodystringrequired

Model name. The value is fixed to omni-flash.

promptbodystringrequired

Video generation prompt. Describe the subject, motion, camera movement, style, and constraints.

Example: "Two women singing and dancing on stage, slow camera push-in, cinematic lighting"

secondsbodystring

Video duration in seconds.

Default: 10
Currently supported value: 10

aspect_ratiobodystring

Output aspect ratio.

Default: 16:9
Available values: 16:9, 9:16

resolutionbodystring

Output resolution.

Default: 720p
Available values: 720p

input_videobodystring

Reference video URL. When provided, the task uses video-reference generation mode.

Only public URLs are supported. data:video/...;base64,... is not supported.

first_image_urlbodystring

Public first-frame image URL. It can be used alone for image-to-video or together with last_image_url for first/last-frame generation. It cannot be used with images or referenceImages.

last_image_urlbodystring

Public last-frame image URL. It must be used with first_image_url and cannot be sent alone. It cannot be used with images or referenceImages.

imagesbodyarray

Public reference image URL array. It cannot be used with first_image_url or last_image_url.

Legacy Compatible Fields

The following legacy fields are still accepted for compatibility, but new integrations should use the fields above:

duration: equivalent to seconds
ratio: equivalent to aspect_ratio
referenceImages: equivalent to images

Modes

Text to Video

Provide model, prompt, seconds, aspect_ratio, and resolution.

First/Last Frame

Use first_image_url, or use both first_image_url and last_image_url.

Multi-image Reference

Use images to provide reference images for subject, style, or visual constraints.

Video Reference

Use input_video to provide a reference video. The system automatically uses video-reference mode.

Constraints

Key constraints:

model only supports omni-flash
seconds currently only supports 10
resolution only supports 720p
aspect_ratio only supports 16:9 and 9:16
input_video does not support base64
last_image_url must be used with first_image_url
images / referenceImages cannot be used with first_image_url / last_image_url
Whether input_video can be combined with image fields is determined by the upstream provider

Material And Content Limits

Category	Item	Requirement
Parameters	Video duration	Recommended reference video input duration is `8-10` seconds. If it exceeds `10` seconds, the upstream provider may process only the first `10` seconds.
Parameters	Material count	Each task supports only `1` reference video and up to `5` auxiliary images. Keeping images within `3` is recommended.
Parameters	Output duration	In video-reference mode, the generated video duration usually follows the uploaded reference video duration. For example, an `8` second reference video usually produces an approximately `8` second output. The current API parameter `seconds` remains fixed to `10`.
Prohibited	Copyrighted elements	Do not use copyrighted elements.
Prohibited	Known IP / characters	Do not use known IP, anime characters, or film/TV characters.
Prohibited	Real celebrity portraits	Do not use real celebrity likenesses.
Prohibited	Brand marks	Do not use famous brand logos, iconic clothing, or signature props.
Prohibited	Protected audio	Do not use protected background music, songs, or specific character voices.
Prohibited	Safety review	Do not include suggestive, violent, sensitive, or otherwise policy-violating content.
Prohibited	Language and recognition	Dialect recognition is not supported. Use standard Mandarin, English, or another clearly spoken language.
Recommended	Material originality	Use original materials that you own, or copyright-free materials that you are authorized to use.
Recommended	Prompt writing	Avoid mentioning specific IP, celebrities, or brands. Avoid instructions such as “make it similar to XX character”. Describe original fictional characters and scenes instead.
Note	File size	Keep video files strictly within `15MB`; images should also be kept within `15MB`.

Response Parameters

idstring

Unique identifier for the video generation task, used for subsequent status queries.

objectstring

Object type, always video.

createdinteger

Task creation timestamp.

modelstring

Model name used for generation. The value is omni-flash.

statusstring

Task status:

queued - Queued
processing - Processing
completed - Completed
failed - Failed

errorstring

Error message returned when the task status is failed.