WAN3 API

Accurate request formats for task creation, reference media, polling and downloads.

Connection

Item	Value
Base URL	`https://video.1route.dev`
Auth	`Authorization: Bearer YOUR_API_KEY`
Body	`application/json`
Models	`GET /v1/models`
Create	`POST /v1/videos`
Query	`GET /v1/videos/{id}`

Use /v1/videos for new integrations. Do not mix legacy and current paths in one client.

List models

curl "https://video.1route.dev/v1/models" \
  -H "Authorization: Bearer YOUR_API_KEY"

Use only model IDs returned for the current key.

Text to video

{
  "model": "wan3.0-video-480p",
  "prompt": "A red kite floats above a green field. The camera slowly moves forward.",
  "seconds": "2",
  "aspect_ratio": "16:9"
}

Image to video

{
  "model": "wan3.0-image-480p",
  "prompt": "The person in image 1 blinks and turns their head slightly. Keep identity consistent.",
  "seconds": "2",
  "aspect_ratio": "9:16",
  "reference_images": [
    {"url": "https://media.example.com/person.jpg"}
  ]
}

First and last frames use role:

{
  "model": "wan3.0-video-720p",
  "prompt": "Transition naturally from the first frame to the last frame.",
  "seconds": "5",
  "aspect_ratio": "adaptive",
  "reference_images": [
    {"url": "https://media.example.com/start.jpg", "role": "first_frame"},
    {"url": "https://media.example.com/end.jpg", "role": "last_frame"}
  ]
}

Up to 10 images are documented. Roles: reference_image, first_frame, last_frame.

Image, video and audio references

{
  "model": "wan3.0-video-480p",
  "prompt": "Keep the subject from image 1, follow motion in video 1 and rhythm in audio 1.",
  "seconds": "10",
  "aspect_ratio": "9:16",
  "reference_images": [{"url": "https://media.example.com/character.jpg"}],
  "reference_videos": [{"url": "https://media.example.com/motion.mp4", "duration": 5}],
  "reference_audios": [{"url": "https://media.example.com/rhythm.mp3"}]
}

This is billed as 15 seconds: 10 output + 5 reference. Image-family models reject video references.

Fields

Field	Type	Meaning
`model`	required string	WAN3 ID returned by the model list
`prompt`	required string	Description; refer to media by order
`seconds`	required string	2–30 whole output seconds
`aspect_ratio`	string	`16:9`, `9:16`, `1:1`, `adaptive`
`reference_images`	object[]	URL and optional role; up to 10
`reference_videos`	object[]	URL and explicit whole duration/seconds
`reference_audios`	object[]	Audio URL
`size` / `resolution`	string	Compatibility fields; model suffix controls the tier

Send seconds only; do not send conflicting duration aliases. Media must be directly downloadable. Prefer HTTPS.

Advanced input.media types are first_frame, last_frame, reference_image, reference_audio, and reference_video. Prefer the top-level reference fields for new clients. Do not duplicate one video across both representations.

Create

curl -X POST "https://video.1route.dev/v1/videos" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  --data '{
    "model": "wan3.0-image-480p",
    "prompt": "The subject in image 1 looks up naturally.",
    "seconds": "2",
    "reference_images": [{"url": "https://media.example.com/reference.jpg"}]
  }'

Persist the full returned id and X-Oneapi-Request-Id. Never automatically retry creation.

Poll

curl "https://video.1route.dev/v1/videos/task_EXAMPLE" \
  -H "Authorization: Bearer YOUR_API_KEY"

Poll about every 15 seconds. Continue for queued and in_progress; completed is success; failed and expired are terminal.

Read the exact metadata.url from a completed response and download immediately. A signed link needs no extra API key. See Tasks and downloads.

Complete Python client

Download wan3_client.py.

It verifies models and prints a request by default. Creation requires both --create and --confirm-cost. Resume without creating by passing --task-id task_....

export WAN3_API_KEY="sk-your-key"
python wan3_client.py --model wan3.0-video-480p --seconds 2
python wan3_client.py --model wan3.0-video-480p --seconds 2 \
  --prompt "A red kite above a green field" --create --confirm-cost
python wan3_client.py --task-id task_EXAMPLE

On this page