AI generated supertall tower in Manhattan, 430m vertical structure sculpted by computational subtraction process, massive carved-out cavities
Wan3 AIWan3 AIUse Wan 3.0 for single takes up to 30 seconds with sound generated in the same pass, or Wan 2.6 for 5, 10 or 15-second clips. Start from a prompt or an image and see the credit cost before you submit.
Open any prompt in the generator above, then change the subject, the camera or the sound to make it yours.
AI generated supertall tower in Manhattan, 430m vertical structure sculpted by computational subtraction process, massive carved-out cavities
HANDHELD GUERILLA STYLE FOUND FOOTAGE VIDEO. GRITTTY AND NARROW ANGLE. HE IS HOLDING THE IPHONE CAMERA AND IS COMPLETELY OUTSIDE THE FRAME, NO GRAPHIC OVERLAY. FIRST PERSON CAMERA VIEW ONLY
Young future astronauts in Space Scout vests and dark heather grey clothing floating and having fun in the zero gravity chambers at Space Camp Huntsville. Excited, joyful, whimsical children's ebook style, inspirational atmosphere.
Under the gaze of the full moon, she rises. Stained glass and starlight, where the last colors of sunset still linger in the sky.
One garden male gnome reaches and touches the other garden female gnome hand and the sky bursts into the stars and the Milky Way
A futuristic digital illustration of BlueGoo and Vera, Mother and Daughter Love is Always the Answer
Preview clips come from our prompt library and were not all rendered with Wan; your Wan 3.0 or Wan 2.6 result will look different.
Wan (Tongyi Wanxiang) is the family of video generation models built by Alibaba's Tongyi Lab. Every version turns a written prompt or a still image into a short video clip; newer versions add longer takes, sound and finer control over frames and references.
Wan3 AI runs two generations. Wan 3.0 entered public beta on 6 August 2026 and became generally available on 24 August 2026; it renders 2 to 30 seconds in one pass, with dialogue, ambience and effects generated together with the picture. Wan 2.6, announced on 16 December 2025, makes fixed 5, 10 or 15-second clips at 720p or 1080p.
Wan3 AI is an independent site, not an Alibaba product. We call the models through a third-party API and show the credit cost of every generation before you submit it.
Facts checked on 9 October 2026 against Alibaba's documentation.
Everything below is switched on in the generator on this page.
Wan 3.0 renders a single continuous clip of 2 to 30 seconds, so a scene can open, turn and resolve without stitching fragments together.
Wan 3.0 creates dialogue, ambience and sound effects in the same pass as the video, so lips, footsteps and impacts land on the right frame. Audio is on by default; switching it off does not change the price.
Upload an opening frame and an optional closing frame and Wan 3.0 plans the motion between them. Or switch to reference mode and give it up to 9 images of a face, an outfit or a product to keep consistent.
Wan 2.6 makes 5, 10 or 15-second clips at 720p or 1080p from a prompt or a first frame. Pick it when you want its look for a short shot; on this site it runs without an audio track.
The generator is at the top of this page; there is nothing to install.
Wan 3.0 is selected by default; switch to Wan 2.6 in the model menu. Start from a prompt, or upload an image to animate as the first frame.
Describe who is in frame, what they do, one camera move and, on Wan 3.0, what we should hear. Choose length, resolution and aspect ratio; the credit cost updates before you submit.
Longer and higher-resolution clips take longer to render. Change one thing at a time between versions. Failed generations are refunded automatically.
The three video models most people choose between on Wan3 AI, with the limits and prices that apply on this site.
| On Wan3 AI | Wan 3.0 | Wan 2.6 | Seedance 2.0 |
|---|---|---|---|
| Clip length | 2–30 s | 5, 10 or 15 s | 4–15 s |
| Resolution | 480p, 720p, 1080p | 720p, 1080p | 480p up to 4K |
| Audio | Generated with the video, on by default | No audio option here | Optional, off by default |
| Image input | First/last frame or up to 9 reference images | First frame | First/last frame, up to 9 images and 3 videos |
| 5 s at 720p | 28 credits | 30 credits | 86 credits |
| Developer | Alibaba Tongyi Lab | Alibaba Tongyi Lab | ByteDance |
| Best for | Long takes with dialogue and sound | Short fixed-length clips | Reference-heavy, directed shots and 4K |
Credits as charged on Wan3 AI on 9 October 2026; the generator shows the exact cost before you submit. Seedance 2.0 also comes in cheaper Fast and Mini versions.
Typical jobs where a Wan clip replaces a shoot or a stock-footage search.
Stage a complete scene with consistent characters and spoken lines in one Wan 3.0 take, instead of cutting short fragments together.
Animate a product photo as the first frame and set the packshot as the last frame, so the clip ends exactly on your product.
Render vertical 9:16 clips for TikTok, Reels and Shorts, with sound generated on Wan 3.0.
Show a process, a place or a historical scene as a short narrated clip for a lesson or a presentation.
Habits that save regenerations, based on Alibaba's prompt guidance and the credit costs on this site.
Start with who or what is in frame, then where it is, then what moves. Add camera, style and sound once the basic shot works.
"Slow dolly in from a medium shot to a close-up" works better than stacking pans, zooms and orbits. Write "fixed shot" if the camera should stay still.
Thirty seconds needs change. Split the clip into timed beats of a few seconds each (0–6 s, 6–12 s …) and say how the shot ends.
Quote dialogue word for word, name the ambience, and add "no background music" if you want none; otherwise Wan 3.0 may add its own.
In reference mode, call uploads Image 1, Image 2 … in upload order and say what each one is for: the face, the outfit, the product.
A 5-second Wan 3.0 draft costs 14 credits at 480p and 55 at 1080p. Lock motion and timing at 480p, then render the final at 1080p.
Release dates, prices, sound, licensing and how this site relates to Alibaba.
Wan (Tongyi Wanxiang) is Alibaba's family of video generation models, built by its Tongyi Lab. Wan3 AI offers two generations: Wan 3.0, which entered public beta on 6 August 2026 and became generally available on 24 August 2026, and Wan 2.6, which Alibaba announced on 16 December 2025.
Wan 3.0 generates 2 to 30 seconds at 480p, 720p or 1080p with audio in the same pass and supports first/last frames and up to 9 reference images. Wan 2.6 offers 5, 10 or 15-second clips at 720p or 1080p for text-to-video and image-to-video.
Start with Wan 3.0: it allows longer takes, sound and frame control, and per second it costs less than Wan 2.6 at 720p on Wan3 AI. Use Wan 2.6 if you prefer its look for a fixed 5, 10 or 15-second clip.
Wan 3.0: a 5-second clip costs 14 credits at 480p, 28 at 720p and 55 at 1080p; 30 seconds at 1080p costs 329. Wan 2.6: a 5-second clip costs 30 credits at 720p and 45 at 1080p; 15 seconds at 1080p costs 133. The exact cost is shown before you generate.
Yes. Both Wan 3.0 and Wan 2.6 accept an image as the first frame. Wan 3.0 also takes an optional last frame, or up to 9 reference images instead of frames.
Wan 3.0 generates dialogue, ambience and sound effects together with the video; audio is on by default. On Wan3 AI, Wan 2.6 clips are generated without an audio option.
Older Wan releases such as Wan 2.1 and Wan 2.2 were published with open weights. Wan 3.0 is a closed model available only through cloud APIs, so there are no official Wan 3.0 weights to download.
No. Wan3 AI is an independent platform and is not affiliated with Alibaba. We run the Wan models through a third-party API provider. Alibaba's own channels are wan.video and Alibaba Cloud Model Studio.
Yes, with a paid plan or credit pack: any purchase gives you a commercial license for all content in your account. Content made on the free plan is for personal use. You are responsible for rights to people, brands or music that appear in your videos.
Yes, Wan 3.0 and Wan 2.6 are on the free plan and limited only by your credits. New accounts in eligible regions get 20 sign-up credits, enough for one 5-second Wan 3.0 video at 480p (14 credits), plus daily check-in credits; longer or higher-resolution clips need a credit pack or plan. Failed generations are refunded automatically. Downloading videos needs a paid plan.
Pick Wan 3.0 or Wan 2.6, write one clear shot and check the credit cost before you generate.