Create FLUX 3 Video videos

Create FLUX 3 Video video tasks via unified POST /v1/videos. Model capabilities are subject to the actual callable range of the current Tikway gateway.

Model	Resolution	Duration	Supported Modes	Audio	Reference Limitations
flux-3-video	HD / FHD (unified field mapping 480p/720p → HD, 1080p → FHD)	5–20 seconds or auto	Vincent, 1–10 keyframes, video continuation, draft enhancement	Supported, enabled by default	Normal audio/video style reference not supported

The difference between the official website capabilities and the current unopened capabilities of the gateway can be found in this audit report.

POST https://api.tikway.ai/v1/videos

cURL

Text to video

curl --request POST \
  --url https://api.tikway.ai/v1/videos \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "flux-3-video",
  "prompt": "A golden retriever runs on the beach at sunset. The camera follows smoothly, and the picture has a natural and delicate film quality.",
  "duration": 8,
  "resolution": "1080p",
  "aspect_ratio": "16:9",
  "generate_audio": true
}
'

202

{
  "id": "9552de3d-171f-4cca-b492-7baa56523239",
  "status": "pending",
  "polling_url": "https://api.tikway.ai/v1/videos/9552de3d-171f-4cca-b492-7baa56523239"
}

Authorizations

Authorization · string · header · required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json

Create a FLUX 3 Video video task. Parameters are narrowed according to the current Tikway gateway capabilities; for differences between models, see operation description and x-model-capabilities.

model · enum<string> · required

FLUX 3 Video Model ID.

Available options: flux-3-video

prompt · string

Video generation or editing prompt words.

Minimum string length: 1

duration · integer enum<string>

An integer of 5–20 seconds, or auto to let the model determine the duration based on content.

Required range: 5 <= x <= 20

aspect_ratio · enum<string>

Video aspect ratio. When producing videos, some models will automatically determine the proportions based on the input material.

Available options: auto, 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, 9:16

resolution · enum<string>

Output resolution. Available values ​​for different models or modes are shown in the model capabilities table.

Available options: 480p, 720p, 1080p

frame_images · object[]

First frame/last frame shortcut. For more keyframes, please pass 1–10 reference_images via input_references, or specify keyframes with time points in extra_body.keyframes.

Maximum array length: 2

Hide child attributes

frame_images. type · enum<string> · required

Reference frame position.

Available options: first_frame, last_frame

frame_images. image_url · string<uri> · required

Reference image URL string.

Minimum string length: 1

input_references · object[]

Tusheng Video can transfer 1–10 reference_images as key frames; video continuation must and can only transfer 1 source_video or first_clip.

Maximum array length: 10

Option 1

Option 2

Hide child attributes

input_references. image_url · string<uri> · required

Image URL string.

Minimum string length: 1

input_references. type · enum<string>

Optional. Material type; if provided, must be consistent with the corresponding URL field.

Available options: image_url, video_url

input_references. role · enum<string>

Material usage. When omitted, normal reference material is inferred from the URL field.

Available options: reference_image, source_video, first_clip

input_references. video_url · string<uri>

Video URL string.

Minimum string length: 1

generate_audio · boolean

Whether to generate synchronized audio, the default is true.

callback_url · string<uri>

User task callback address. Called by Tikway after the task status changes, it is not directly passed to the model upstream.

Pattern: ^https://

extra_body · object

FLUX 3 Video native extension parameters.

Hide child attributes

extra_body. mode · enum<string>

Available options: text_to_video, image_to_video, video_to_video, draft_enhance

extra_body. keyframes · string any[]

1–10 images, or timed keyframes arranged by [seconds, image URL].

extra_body. start_video · string<uri>

Video continuation input.

extra_body. draft_cache · string

Cache URL or Base64 required for draft enhancement.

extra_body. version · enum<string>

Available options: latest

extra_body. safety_tolerance · integer · default: 2

Required range: 0 <= x <= 4

extra_body. draft · boolean · default: false

Whether to generate a low-cost HD draft first.

Response

202 - application/json

POST /v1/videos Response parameters after the video task is successfully created. The interface returns HTTP 202.

id · string · required

The unique identifier of the video generation task, used for subsequent query of task status.

Minimum string length: 1

Example:

"9552de3d-171f-4cca-b492-7baa56523239"

status · enum<string> · required

Task creation status. pending means that the task has been created and is waiting for submission; queue means that the task has entered the generation queue.

Available options: pending, queue

Example:

"queue"

polling_url · string<uri> · required

Video task query address, the client can poll the task status through GET request.

Pattern: ^https://

Example:

"https://api.example.com/v1/videos/9552de3d-171f-4cca-b492-7baa56523239"