NewWan 3.0 is live on imageatWan 3.0
30 seconds. 1080p. With sound.
Alibaba's newest video model makes up to 30 seconds of 1080p video in a single take, with dialogue, ambience and sound effects generated alongside the picture. From text, from a photo, or between two frames.
Every video on this page was generated with Wan 3.0 on imageat, sound included
One prompt. One long take.
Wan 3.0 renders the whole clip in a single generation instead of stitching short pieces together. A camera move can start in one place and end somewhere else entirely, and the scene stays consistent along the way. This 15-second drone flight was one prompt and one run.
Prompt: One continuous drone shot, no cuts. The camera starts low over turquoise water beside a white Mediterranean fishing village, glides past moored wooden boats and a harbor cafe, rises up the hillside over terracotta rooftops and a church bell tower, and finally reveals the open sea and a sailboat on the horizon at sunset. Smooth steady flight, warm golden light. Seagulls, gentle waves, a distant church bell.
Sound made with the picture
Dialogue, room tone and the sounds of the action come out of the same generation, so the sizzle hits when the patty does and the words line up with the speaker. Put a line in quotes and Wan 3.0 says it. Turn the sound on below.
Dialogue and lip sync. Vertical selfie-style video. A friendly woman in her late twenties with curly hair and a denim jacket talks straight to camera in a bright coffee shop, holding a latte. She smiles and says: "Okay, I finally found the best coffee in the city, and you have to try the oat milk flat white." Natural handheld phone framing, soft window light, background chatter and espresso machine hiss.
Sound that matches the action. Overhead close shot of a smash burger patty hitting a hot cast-iron griddle, a spatula presses it flat, edges crisp and sizzle, a slice of cheddar is laid on top and starts to melt, steam rising. Restaurant kitchen at night, warm tungsten light, grease droplets catching the light. Loud sizzle and the scrape of the spatula.
Built for real shots
Smoother motion, stronger scene fidelity and better visual coherence than earlier Wan versions, which shows most in the hard cases: bodies in motion, close product detail and vertical framing.
Start Now!Native vertical
9:16 is generated as a vertical frame, not cropped from landscape, so faces and products stay centered for TikTok, Reels and Shorts.
Full-body motion
Fast footwork, spins and freezes stay readable: limbs, weight and ground contact hold together through the whole take.
Detail at 1080p
Leather grain, brushed metal and hands handling a product, sharp enough for an ad without upscaling.
Start from your own image
Upload a photo, render or illustration and describe what should move. Wan 3.0 keeps it as the opening frame. Add a last frame too and it directs the shot from one image to the other.
Your imageMotion prompt: Rain keeps falling, steam drifts up from the ramen bowls, the lanterns sway gently in the wind, a person with a clear umbrella walks past the shop window, the camera slowly pushes in toward the counter. Rain on the awning, quiet city ambience.
How to use Wan 3.0
Describe the shot
Write the subject, the action, the camera move and the sound. Or upload a photo to use as the first frame.
Set length and format
Pick 2 to 30 seconds, 480p to 1080p and an aspect ratio. Leave audio on to get sound with the picture.
Generate and download
Wan 3.0 renders the clip in one pass. Download the MP4 from your history, or iterate on the prompt.
Everything Wan 3.0
can do on imageat
| Modes | Text to video, image to video, first and last frame, reference images (API) |
|---|---|
| Length | 2 to 30 seconds, any whole second, rendered as one take |
| Resolution | 480p, 720p or 1080p |
| Aspect ratio | 16:9, 9:16, 1:1, 4:3, 3:4, or Auto to follow the prompt or image |
| Audio | Dialogue, ambience and sound effects generated in the same pass, on by default |
| References | Up to 10 images through the API to keep a character, product or place consistent |
| Output | MP4, 30 fps, saved to your history with the prompt |
Write prompts like a shot list
Frame it
Say how the shot is framed and moves: close-up, wide, low angle, tracking, slow dolly-in, drone.
Light it
Name the light and the look: golden hour, neon, tungsten, soft window light, 35mm film.
Sequence it
Describe the action in order. Wan 3.0 follows a sequence of beats across a long take.
Hear it
Write the sound: the line someone says in quotes, the sizzle, the crowd, the wind.
Made with Wan 3.0
Every prompt is below its clip. Open one in the studio to make your own version.
15-second drone take
One continuous drone shot, no cuts. The camera starts low over turquoise water beside a white Mediterranean fishing village, glides past moored wooden boats and a harbor cafe, rises up the hillside over terracotta rooftops and a church bell tower, and finally reveals the open sea and a sailboat on the horizon at sunset. Smooth steady flight, warm golden light. Seagulls, gentle waves, a distant church bell.
Try this promptTalking to camera
Vertical selfie-style video. A friendly woman in her late twenties with curly hair and a denim jacket talks straight to camera in a bright coffee shop, holding a latte. She smiles and says: "Okay, I finally found the best coffee in the city, and you have to try the oat milk flat white." Natural handheld phone framing, soft window light, background chatter and espresso machine hiss.
Try this promptProduct close-up
Premium product commercial. Close-up of hands slowly unboxing a matte black wireless headphone case on a walnut desk, lifting the headphones out and turning them in the light. Macro detail on leather ear cushions and brushed aluminum. Shallow depth of field, soft key light with a warm rim light, slow dolly-in. Quiet room tone, the soft click of the case opening.
Try this promptOn the griddle
Overhead close shot of a smash burger patty hitting a hot cast-iron griddle, a spatula presses it flat, edges crisp and sizzle, a slice of cheddar is laid on top and starts to melt, steam rising. Restaurant kitchen at night, warm tungsten light, grease droplets catching the light. Loud sizzle and the scrape of the spatula.
Try this promptGolden-hour fashion
Fashion film, vertical. A model in a flowing ivory linen dress walks along a windy clifftop path at golden hour, the fabric and her hair moving in the wind, she turns her head toward the camera and keeps walking. Tracking shot at waist height, warm low sun, lens flare, 35mm film look. Wind and distant waves.
Try this promptBreakdance at dusk
A breakdancer in a red tracksuit performs on a concrete basketball court at dusk: toprock footwork, drops into a windmill, then freezes in a one-hand handstand. Full body always in frame, wide shot, low angle, camera slowly circles. Streetlights flicker on, a small crowd cheers, boombox hip-hop beat.
Try this promptA still, brought to life
Rain keeps falling, steam drifts up from the ramen bowls, the lanterns sway gently in the wind, a person with a clear umbrella walks past the shop window, the camera slowly pushes in toward the counter. Rain on the awning, quiet city ambience.
Try this promptNeon alley in the rain
Cinematic tracking shot of a young woman in a flowing red coat walking through a neon-lit Tokyo alley at night, light rain, reflections on wet pavement, steam rising from food stalls, soft city ambience
Try this promptPay per second
Audio included at no extra cost. A 5-second 720p clip is 20 credits, a 10-second 1080p clip is 70. One credit is $0.10, and failed videos are refunded automatically.
A Wan 3.0 API, one key
From $0.20 per video. Text, image, first/last frame and up to 10 reference images. Also available to AI agents over MCP.
curl -X POST "https://api.imageat.com/v1/videos" \
-H "Authorization: Bearer iat_live_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "wan3",
"prompt": "a drone shot over a harbor at sunset",
"duration": "10s",
"resolution": "1080p",
"aspectRatio": "16:9"
}'
# → { "status": "pending", "taskId": "..." }Make your first
Wan 3.0 video
Wan 3.0 sits in the imageat video studio next to Seedance, Kling and Veo. Pick a length, write the shot, and pay only for the seconds you generate.
Questions, answered
Wan is developed by Alibaba's Tongyi Lab. imageat is not affiliated with Alibaba.