Real outputs from the public examples of this model on Replicate, with the prompt and settings that produced them. They were not generated live on this page.
Prompt
A sports car is driving very fast along a beach at sunset
Alibaba's Tongyi Lab in Hangzhou runs the Qwen LLM family and the Wanxiang generative-media family. After Wan 2.0 (mid-2024) and Wan 2.1 (early 2025), the team released Wan 2.2 in 2025 as the next-generation open-weight video model. Wan 2.2 ships as purpose-tuned variants for Text-to-Video, Image-to-Video and Audio/A2V. Wan 2.2 Text-to-Video is the flagship pure-text-conditioned variant and replaces Wan 2.1 T2V-14B as the principal open-weight text-to-video reference for the Chinese research community. The Wan team consistently rank near the top of open-model VBench leaderboards and ship reproducible training code under a permissive Wan-series licence.
Diffusion Transformer (DiT) with 3D causal VAE; MoE-style scaling
Wan 2.2 Text-to-Video is a Diffusion Transformer operating on a 3D causal Wan-VAE latent. Wan 2.2 introduces architectural refinements over Wan 2.1: improved 3D Rotary Position Embeddings, larger attention windows, and (in the flagship) a Mixture-of-Experts feed-forward design that routes tokens to specialist experts. Text conditioning uses a Qwen-family multilingual encoder with strong Chinese-English capability. The denoiser is trained with Flow Matching on a curated multi-million-clip multilingual video corpus with synthetic dense bilingual captions. Native generation is 5 seconds at 720p / 24 fps (with 1080p extensions). The training recipe and weights are open-source on Hugging Face and GitHub under the Wan-series permissive licence, designed to enable broad commercial and research use.
Bilingual Chinese/English prompts via Qwen-based text encoder
MoE-style scaling and improved 3D RoPE in flagship variant
Permissive Wan-series licence for research and commercial use
Top-tier results on VBench among open-weight models
Active community ecosystem (LoRAs, fine-tunes, ComfyUI nodes)
Reproducible training recipe and code
Best for: open-source video pipelines, research, on-prem creative tooling, branded fine-tunes.
Entraînement et licence
Curated multi-million-clip multilingual video corpus filtered for aesthetics, motion and caption quality, with dense bilingual captions; specifics documented in Wan technical materials.
Licence: Open weights under an Apache-style permissive licence (Wan-series release).
curl https://railwail.com/api/v1/videos/generations \
-H "Authorization: Bearer $RAILWAIL_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "wan-t2v",
"prompt": "A slow drone shot over a misty pine forest at sunrise"
}'
# Response: {"job_id": "...", "status": "queued", ...}
# Poll until status is completed, failed or cancelled:
curl https://railwail.com/api/v1/jobs/JOB_ID \
-H "Authorization: Bearer $RAILWAIL_API_KEY"
import os
import time
import requests
API = "https://railwail.com/api/v1"
headers = {"Authorization": f"Bearer {os.environ['RAILWAIL_API_KEY']}"}
job = requests.post(
f"{API}/videos/generations",
headers=headers,
json={
"model": "wan-t2v",
"prompt": "A slow drone shot over a misty pine forest at sunrise",
},
).json()
while True:
status = requests.get(f"{API}/jobs/{job['job_id']}", headers=headers).json()
if status["status"] in ("completed", "failed", "cancelled"):
break
time.sleep(5)
print(status["status"], status.get("output_url"))
const API = "https://railwail.com/api/v1";
const headers = {
Authorization: `Bearer ${process.env.RAILWAIL_API_KEY}`,
"Content-Type": "application/json",
};
const job = await fetch(`${API}/videos/generations`, {
method: "POST",
headers,
body: JSON.stringify({
model: "wan-t2v",
prompt: "A slow drone shot over a misty pine forest at sunrise"
}),
}).then((r) => r.json());
let status;
do {
await new Promise((r) => setTimeout(r, 5000));
status = await fetch(`${API}/jobs/${job.job_id}`, { headers }).then((r) => r.json());
} while (!["completed", "failed", "cancelled"].includes(status.status));
console.log(status.status, status.output_url);
MiniMax Hailuo 02 on Replicate. Text-to-video and image-to-video producing 6s or 10s clips at 768p standard or 1080p pro. Known for accurate real-world physics and stable motion.
