Wan3.0 Video Generation

Wan3.0 (wan3.0-video) is Alibaba Tongyi Wanxiang's Wan3.0 video generation model. It supports text-to-video, image-to-video (first frame / reference image), video reference and video editing, up to 30 seconds at 30fps. It uses the unified POST /v1/videos endpoint and returns an OpenAI Video style response (id / status / metadata.url), sharing the same task polling mechanism as other video models.

Billing

Wan3.0 is billed by actual generated duration (seconds), tied to the resolution tier (480P / 720P / 1080P have different unit prices). Exact unit prices and billing expressions are shown on the corresponding model card in the console "Model Plaza" (prices are intentionally omitted from this documentation).

Endpoints

• Create task: POST $URL/v1/videos
• Query task: GET $URL/v1/videos/{task_id}

Compatible alias

An OpenAI Sora-style alias is also supported: POST $URL/v1/video/generations, GET $URL/v1/video/generations/{task_id} — equivalent to the above.

Supported Models

Model name	Positioning	Resolutions
wan3.0-video	Standard edition, all-in-one reference generation	480P / 720P / 1080P
wan3.0-video-prime	High-speed edition, significantly faster end-to-end, capability aligned with standard	480P / 720P / 1080P

Within the same model name, the resolution tier is selected via the size request parameter (see "Parameters" below).

Parameters

Parameter	Type	Required	Description
model	string	yes	wan3.0-video / wan3.0-video-prime
prompt	string	yes	Video description; recommend specifying subject, action, scene, and camera movement
size	string	no	Resolution tier: 480p / 720p / 1080p, defaults to 1080P
seconds	string	no	Output duration in seconds, e.g. "5", "10"; range 2–30 without video input, "-1" for smart duration
image	string	no	Reference image URL (single, used as the first frame for image-to-video)
images	array	no	Reference image URL array (multi-image reference)
input_reference	string	no	Reference / to-be-edited video URL, for video reference generation or video editing
metadata.resolution	string	no	Advanced override: directly specify upstream resolution 480P / 720P / 1080P
metadata.duration	number	no	Override output duration (seconds)
metadata.seed	integer	no	Random seed for reproducibility

Reference material

Publicly accessible image/video URLs are supported (the upstream server fetches them; intranet or local addresses will fail). Wan3.0 does not provide a standalone asset library — pass reference material via URL.

Request Examples

Text-to-video

curl

# 1. Create video task (720P, 5 seconds)
curl $URL/v1/videos \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video",
    "prompt": "A cat wearing sunglasses walking down a neon street, cyberpunk style, camera slowly pushing in",
    "seconds": "5",
    "size": "720p"
  }'

# 2. Query task status (replace {task_id} with the id returned above)
curl $URL/v1/videos/{task_id} \
  -H "Authorization: Bearer $YOUR_API_KEY"

Python

import time
import requests

url = "$URL/v1/videos"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json",
}
payload = {
    "model": "wan3.0-video",
    "prompt": "A cat wearing sunglasses walking down a neon street, cyberpunk style",
    "seconds": "5",
    "size": "720p",
}

task = requests.post(url, json=payload, headers=headers).json()
task_id = task["id"]

while True:
    result = requests.get(f"{url}/{task_id}", headers=headers).json()
    if result["status"] in ("completed", "failed"):
        break
    time.sleep(8)

print(result)

Image-to-video (first frame)

curl $URL/v1/videos \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video",
    "prompt": "Make the person in the frame turn and walk toward the camera",
    "image": "https://example.com/ref.png",
    "seconds": "5",
    "size": "1080p"
  }'

Video reference / video editing

curl $URL/v1/videos \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video",
    "prompt": "Convert the entire scene into a claymation style",
    "input_reference": "https://example.com/clip.mp4",
    "seconds": "5",
    "size": "720p"
  }'

High-speed edition (prime)

Just change model to wan3.0-video-prime; all other parameters are identical:

curl $URL/v1/videos \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video-prime",
    "prompt": "Waves crashing against rocks, slow motion, cinematic",
    "seconds": "5",
    "size": "1080p"
  }'

Response Fields

The create and query endpoints return the same structure (OpenAI Video style):

Field	Type	Description
id	string	Task ID
status	string	Task status (see below)
metadata.url	string	Video URL after generation (returned only when completed)
progress	integer	Generation progress, 0~100
completed_at	integer	Completion timestamp (seconds, returned only when completed)
error.message	string	Failure reason (returned only when failed)

Task Status

Status	Meaning	Suggested action
queued	Queued	Keep polling
in_progress	Generating	Keep polling
completed	Success	Get video URL from metadata.url; download and store promptly
failed	Failed	Check error.message

Poll every 5~10 seconds

Download and store the video as soon as it completes: the URL is time-limited and becomes inaccessible after expiry.

Response Example

{
  "id": "task_xxxxxxxx",
  "status": "in_progress",
  "progress": 30
}

After completion:

{
  "id": "task_xxxxxxxx",
  "status": "completed",
  "progress": 100,
  "metadata": {
    "url": "https://cdn.example.com/.../result.mp4"
  },
  "completed_at": 1786431715
}

FAQ

Q: How to write better prompts?

Include four elements: subject (who/what), action (doing what), scene (where), camera (how shot, e.g. push-in / orbit / overhead). Avoid overly abstract descriptions.

Q: How long does a task take?

Depends on duration and resolution. 720P 5s typically finishes within 1~3 minutes; 1080P takes longer. Poll the query endpoint for the final status; tasks not finished within 20 minutes auto-fail.

Q: Do generated video links expire?

Yes. The video URL is time-limited — download and store it to your own storage promptly after completion.

Q: Is there an asset library?

Wan3.0 does not currently provide a standalone asset library. Use publicly accessible image/video URLs as reference frames or reference videos (see image / images / input_reference above).