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.
Input
Prompt
Close-up shot of an elderly sailor wearing a yellow raincoat, seated on the deck of a catamaran, slowly puffing on a pipe. His cat lies quietly beside him with eyes closed, enjoying the calm. The warm glow of the setting sun bathes the scene, with gentle waves lapping against the hull and a few seabirds circling slowly above. The camera slowly pushes in, capturing this peaceful and harmonious moment.
Alibaba's Tongyi Lab runs the Qwen LLM family and the Wanxiang generative-media family (image and video). 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 two purpose-tuned variants -- Image-to-Video (Wan-I2V) and Text-to-Video (Wan-T2V) -- alongside an audio/A2V variant. Wan 2.2 Image-to-Video is positioned for animating a user-supplied first frame with strong identity preservation and rich motion, and remains under the permissive Wan-series licence that enables broad commercial and research use.
Diffusion Transformer (DiT) with image-conditioning adapters and 3D causal VAE
Wan 2.2 Image-to-Video is a Diffusion Transformer operating on a 3D causal Wan-VAE latent. The model is initialised from a Wan 2.2 base and fine-tuned for image-to-video with dedicated conditioning adapters that inject features from the user-provided first frame at multiple resolutions of the DiT, ensuring strong identity preservation and continuity. Text conditioning uses a Qwen-family multilingual encoder with strong Chinese-English capability. Native generation is 5 seconds at 720p / 24 fps (with 1080p in extended modes). Wan 2.2's MoE-style scaling and improved 3D RoPE enable better motion physics and reduced identity drift relative to Wan 2.1. The training recipe includes a curriculum mixing image-to-image, image-to-short-video and image-to-long-video data with synthetic dense captions. The release is open-weight on Hugging Face and GitHub under the Wan-series permissive licence.
Paramètres
14 billion (I2V flagship); smaller variants available
Bilingual Chinese/English prompts via Qwen-based text encoder
Identity-preserving conditioning on a user first frame
Permissive Wan-series licence for research and commercial use
Active community ecosystem on Hugging Face / ComfyUI
Compatible with LoRA fine-tunes and reference adapters
Reproducible training recipe documented in technical reports
Best for: open-source image animation pipelines, e-commerce product motion, character animation.
Formation & licence
Curated multi-million-clip multilingual video corpus with paired first-frame conditioning data and dense bilingual captions; specifics documented in the Wan 2.2 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-i2v",
"prompt": "A slow drone shot over a misty pine forest at sunrise",
"start_image": "https://example.com/first-frame.jpg"
}'
# 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-i2v",
"prompt": "A slow drone shot over a misty pine forest at sunrise",
"start_image": "https://example.com/first-frame.jpg",
},
).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-i2v",
prompt: "A slow drone shot over a misty pine forest at sunrise",
start_image: "https://example.com/first-frame.jpg"
}),
}).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.
