20% off all GPT 🎉 30% off DeepSeekLearn more

Wan 3.0

Video Generation
alibaba/wan-3.0

Wan 3.0 is Alibaba's Tongyi Wanxiang video generation model, released on 2026-08-24, unifying text-to-video, image-to-video (first frame or first-and-last frame), and reference-to-video in a single model. Inputs can be mixed across text, image, video, and audio, with up to 10 reference images and 5 reference video or audio clips. It outputs 480P, 720P, or 1080P clips of 2 to 30 seconds across six aspect ratios plus adaptive, emits synchronized audio by default, and supports video continuation and editing. Billed per second of output, by resolution. Accessible via the OpenAI-compatible protocol.

Duration
2-30s
Audio
Supported
Released
2026-08-24
Generation Modes
t2vText to Videoi2vImage to Videov2vVideo to Video
Resolutions
480p720p1080p
Aspect Ratios
16:94:31:13:49:16adaptive
Capabilities
Video Input
Available Providers
AliyunAlibaba Cloud
Supported Protocols
openai

Providers

Aliyunprovider.type: "aliyun"
From $0.0429/s
ResolutionModeAudioPrice / Second
480pAnyAny$0.0429/s
720pAnyAny$0.0857/s
1080pAnyAny$0.1714/s
All other combinations$0.1714/s
Protocols
openai/v1/videos
Alibaba Cloudprovider.type: "alicloud"
From $0.054/s
ResolutionModeAudioPrice / Second
480pAnyAny$0.054/s
720pAnyAny$0.11/s
1080pAnyAny$0.21/s
All other combinations$0.21/s
Protocols
openai/v1/videos

Billed per second of generated video. Supported parameter combinations and prices are determined by the live API and may vary by model version.

Code Examples

import time
import requests
BASE = "https://api.ofox.run/v1/videos"
HEADERS = {"Authorization": "Bearer YOUR_OFOX_API_KEY"}
# Create the generation task
task = requests.post(BASE, headers=HEADERS, json={
"model": "alibaba/wan-3.0",
"prompt": "A golden retriever running on the beach at sunset",
"duration": 2,
"resolution": "1080p",
"generate_audio": True,
"provider": {"type": "aliyun"},
# Or: "provider": {"type": "alicloud"}
}).json()
# Poll until the task reaches a terminal state
while True:
task = requests.get(f"{BASE}/{task['id']}", headers=HEADERS).json()
if task["status"] in ("completed", "failed", "cancelled", "expired"):
break
time.sleep(10)
# Prefer mirror_urls (persistent) then fall back to unsigned_urls (24h)
print((task.get("mirror_urls") or task.get("unsigned_urls") or [None])[0])
Wan 3.0 prompts & examples · 71 casesVerbatim prompts, full parameters and a playable clip, every one cited to its first-party source.

Frequently Asked Questions

Wan 3.0 on Ofox.ai costs from $0.054 per second of generated video. The exact rate depends on resolution, generation mode, and audio. Pay-as-you-go, no monthly fees.