Wan 3.0

Wan 3.0 (alibaba/wan-3.0) is a video generation model from Alibaba for text-to-video, image-to-video, and reference-guided video generation. It produces 480p, 720p, or 1080p video with durations from 2 to 30 seconds. The fast mode with the same geometry is Wan 3.0 Prime (alibaba/wan-3.0-prime): text-to-video and first-frame image-to-video. Generation is asynchronous: POST /v1/videos returns a job, then you poll the status and download the mp4.

Parameters

Field	Description
model	alibaba/wan-3.0 (or the short wan-3.0) or alibaba/wan-3.0-prime
prompt	Scene description, required
seconds	Any integer from 2 to 30; defaults to 2
size	A WxH size from the table below; defaults to 1280x720
input_reference	Image-to-video: one image as the starting frame, {"image_url": "data:image/...;base64,..."}
input_references	Reference-to-video: 2 to 9 images as an array, also data: URLs only
seed	Reproducibility seed: an integer in 0…2147483647. The same prompt with the same seed yields the same clip; without it the upstream picks a random one

Fifteen allowed sizes — five aspect ratios at three resolutions:

Aspect	480p	720p	1080p
16:9	852x480	1280x720	1920x1080
9:16	480x852	720x1280	1080x1920
1:1	480x480	720x720	1080x1080
4:3	640x480	960x720	1440x1080
3:4	480x640	720x960	1080x1440

A value outside the list returns 400 with the allowed options listed — check the sizes field in the catalog when in doubt.

Examples

Text-to-video

Python

import requests

job = requests.post(
    "https://api.mixen.ai/v1/videos",
    headers={"Authorization": f"Bearer {API_KEY}"},
    json={
        "model": "alibaba/wan-3.0",
        "prompt": "A metro train emerges from a tunnel onto a bridge, the city waking up, a long tracking shot",
        "seconds": "12",
        "size": "1280x720",
    },
).json()
print(job["id"], job["status"])

cURL

curl https://api.mixen.ai/v1/videos \
  -H "Authorization: Bearer $MIXEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "alibaba/wan-3.0",
    "prompt": "A metro train emerges from a tunnel onto a bridge, the city waking up, a long tracking shot",
    "seconds": "12",
    "size": "1280x720"
  }'

Image-to-video from a starting frame

The image goes in as a single data: URL — external links are not accepted:

import base64
import requests

with open("street.jpg", "rb") as f:
    img = "data:image/jpeg;base64," + base64.b64encode(f.read()).decode()

job = requests.post(
    "https://api.mixen.ai/v1/videos",
    headers={"Authorization": f"Bearer {API_KEY}"},
    json={
        "model": "alibaba/wan-3.0",
        "prompt": "The camera drifts along the street as passers-by start moving, soft morning light",
        "seconds": "8",
        "size": "1280x720",
        "input_reference": {"image_url": img},
    },
).json()
print(job["id"], job["status"])

Reference-to-video from several images

Two to nine references — the model assembles the scene while keeping the characters and style of the inputs:

{
  "model": "alibaba/wan-3.0",
  "prompt": "The hero from the first frame walks down the alley from the second, neon signs",
  "seconds": "6",
  "size": "1280x720",
  "input_references": [
    "data:image/png;base64,iVBORw0...",
    "data:image/png;base64,iVBORw0..."
  ]
}

From there it is the usual loop: GET /v1/videos/{id} every 5–10 seconds until completed, then GET /v1/videos/{id}/content for the mp4. Full cycle — video guide.

Price

Billing is per second and depends on the resolution. Wan 3.0: 480p — 6.6846 ₽ ($0.075) per second, 720p — 13.3693 ₽ ($0.15), 1080p — 26.7386 ₽ ($0.3). Wan 3.0 Prime: 480p — 9.0911 ₽ ($0.102), 720p — 18.717 ₽ ($0.21), 1080p — 37.434 ₽ ($0.42). You are charged actual duration × the size’s rate. Current prices — catalog.

Strengths and limits

• Durations from 2 to 30 seconds — a continuous range: both short inserts and long takes from one model.
• Five frame formats at three resolutions up to 1080p.
• Three inputs: text, a starting frame, and 2–9 references; Prime’s primary scenario is text and the first frame.
• Prime is the fast mode: the same geometry and durations at a higher rate, for when the clip is needed sooner.
• The first-to-last frame transition is not supported — only the starting frame can be passed.
• The tiers differ only in price and speed — parameters, sizes, and references are shared.

All models — in the catalog. The general video workflow — video guide.