Wan 3
Alibaba's Wan 3 generates up to 30 seconds of 1080p video in a single pass, with audio in the same generation and every aspect ratio from 16:9 to 9:16.
Your prompt opens in the playground, where you generate.
Thirty Seconds, 1080p, Sound in the Same Pass
Up to 30 Seconds in One Pass
Anywhere from 2 to 30 seconds in a single generation, not short clips stitched together afterwards. One continuous pass is what lets camera movement and one-take shot language hold together across a whole scene. Leave the length unset and the model reads the prompt and picks its own.
1080p With Sound
Output at 1080p with audio generated in the same pass as the picture, so dialogue, ambience, and on-screen action land together instead of being dubbed on afterwards. Pick any ratio from 16:9 down to 9:16, or let the model choose one to suit the shot.
Full-Body Motion at Speed
Fast, athletic movement is the hardest thing to keep coherent, and the clip above leads with it: a dance battle at night, shot in a crowd, under mixed neon and streetlight. Limbs, weight, and contact with the ground stay readable through the whole take. Turn on thinking and the model reasons about composition and motion before it renders a frame.
Prompt it with more than a prompt
Text alone is a narrow way to describe a shot. Reference-to-video conditions on what you already have: up to 10 images, 5 video clips, and 5 audio tracks in one request, plus a document or a public web page when thinking is on. Address them positionally in the prompt to say which reference is the character and which is the location.
- Text
- Image
- Video
- Audio
- Documents
- Web pages
Shot for the feed
Native 9:16, generated at portrait ratio rather than cropped down from a landscape frame. Every prompt is shown in full, so you can copy one and run it yourself.
Talking straight to camera
"A woman in her late twenties sits in a bright bedroom holding a mint-green serum bottle with a gold cap up beside her face, talking straight to camera about it, turning the bottle to show the label. Vertical framing, handheld phone camera, soft daylight through window blinds behind her. Natural skin texture, a few loose strands of hair, an unforced smile between sentences. She speaks in a warm conversational voice."
Hands, texture, and packaging
"Two hands open a mint-green matcha latte box on a pale pink tabletop, lift the lid away, then dip fingers into a glass jar of matcha powder and let it fall back through in a fine stream. Vertical framing, overhead and slightly angled, soft diffused studio light. Powder clings to the fingertips, a little dusts the table, the jar glass is slightly smudged. Quiet room tone."
Steam, sizzle, and close detail
"A chef's hands finish a bowl of ramen on a small counter kitchen pass: broth steaming, chopsticks laying two slices of chashu, a ladle of chilli oil breaking the surface into red rings. Vertical framing, handheld 35mm, warm overhead tungsten against a cold window edge, shallow focus. Steam hazes the lens, a thumbprint smudges the bowl rim, the scallion scatter is uneven. Ambient kitchen room tone, the clink of ceramic on steel."
Movement, fabric, and light
"A woman in her late twenties walks toward camera down a narrow city street at golden hour, a long camel coat catching the wind, boots on wet cobblestone. Vertical framing, handheld 50mm follow, backlit with a low flare across the frame. Loose strands of hair cross her face, one boot toe is scuffed, the pavement reflects unevenly. Street ambience with distant traffic."
Generate, animate, or condition on references
Create a shot from a prompt, animate a still with optional control of the last frame, or condition the generation on images, video, audio, and documents you already have.
Wan 3.0 is the latest generation AI video model, delivering enhanced motion smoothness, superior scene fidelity, and greater visual coherence.
Wan 3.0 is the latest generation AI video model, delivering enhanced motion smoothness, superior scene fidelity, and greater visual coherence.
Wan 3.0 is the latest generation AI video model, delivering enhanced motion smoothness, superior scene fidelity, and greater visual coherence.
How to access the Wan 3 API
The client API handles the request submit protocol. It will handle the request status updates and return the result when the request is completed.
import { fal } from "@fal-ai/client";
const result = await fal.subscribe("alibaba/wan-3.0/text-to-video", {
input: {
prompt: "A dancer spins under mixed neon and streetlight, a crowd closing in around her, handheld 35mm",
resolution: "1080p",
duration: 10,
audio: true,
},
logs: true,
onQueueUpdate: (update) => {
if (update.status === "IN_PROGRESS") {
update.logs.map((log) => log.message).forEach(console.log);
}
},
});
console.log(result.data);
console.log(result.requestId);
Common questions about Wan 3
What is Wan 3?
Wan 3.0 is the next generation of the Wan video model family from Alibaba's Tongyi Lab, the team behind Wan 2.1 through Wan 2.7. It generates video from text, from a still image, or from reference material, at up to 30 seconds a shot with audio in the same pass, and it can reason about composition and motion before rendering.
Which Wan 3 endpoints are available on fal?
Three: text-to-video for generating a shot from a prompt, image-to-video for animating a still, with an optional end frame so you can set where the shot finishes, and reference-to-video for conditioning on material you already have. All three share the same duration, resolution, aspect ratio, and audio controls.
What resolutions, durations, and aspect ratios does Wan 3 support?
Output runs at 480p, 720p, or 1080p, with 1080p the default. Durations run from 2 to 30 seconds in a single generation, and you can leave the length unset to let the model choose one from your prompt and reference media. Aspect ratios cover 16:9, 4:3, 1:1, 3:4, and 9:16, or the model can pick the ratio that suits the shot. Audio is generated alongside the video by default.
What can I use as a reference with Wan 3?
Reference-to-video takes up to 10 reference images, up to 5 reference video clips totalling 15 seconds at 16fps or higher, and up to 5 reference audio tracks totalling 15 seconds, in a single request. With thinking turned on it also reads a document you upload or a public web page you link, so a brief or a product page can seed a generation directly. Your prompt can address the references positionally, which is how you say which one is the character and which one is the location.
What does thinking do on Wan 3?
Thinking is an optional mode, off by default, that has the model reason about composition, staging, and motion before it renders a frame. It is worth turning on for prompts that describe a sequence of events rather than a single continuous action, and it is required for the document and web-page reference inputs, since those have to be read and interpreted before they can be turned into a shot.
How much does Wan 3 cost on fal?
Pricing is per second of generated video and varies by output resolution: $0.05/s at 480p, $0.10/s at 720p, and $0.20/s at 1080p. A 5-second clip at 720p costs $0.50, and a 10-second clip at 1080p costs $2.00. The rate is the same on all three endpoints, pay-per-use with no minimums and no subscription. Check each model page for the current rate, which fal notes is subject to change.
Is Wan 3 released as open weights?
Alibaba's published Wan weights currently stop at Wan 2.2 under Apache 2.0, and the releases since then have been commercial API models. Wan 3 is available on fal as an API model. We will note the licensing here if Alibaba publishes weights for it.
Which other Wan models can I use on fal?
Wan 2.7 remains available, with text-to-video, image-to-video, reference-to-video, and instruction-based video editing at 720p and 1080p, native audio, and durations from 2 to 15 seconds. Wan 2.6 is also available, along with open-weight Wan 2.1 and 2.2 endpoints and LoRA trainers.
Can I use Wan models for commercial projects?
Yes. Content generated through the fal.ai API can be used in commercial projects. Check fal.ai's terms of service for full details on usage rights and licensing.
Get in touch about Wan 3
Building something on Wan 3 and need higher rate limits, dedicated capacity, or a hand with the integration? Leave your details and our team will reach out.