Wan AI Video and Image Generator: Every Version
Wan is Alibaba's family of video and image generation models. Zero Kit runs Wan 3.0 for video and Wan 2.7 Image for pictures in one editor.
In Chinese, Alibaba's Wan is called Tongyi Wanxiang.
On Zero Kit the family has two pages:
- Wan 3.0 covers video, in a standard version and a faster Prime version.
- Wan 2.7 Image covers pictures, in a standard version and a Pro version.
Each page has its own sample prompts and a full price table.
Wan 3.0 is what Alibaba calls an all-in-one, reference-based video model.
One model turns a prompt into a clip, starts from a first frame or moves between a first and a last frame, and builds a clip from a mix of reference images and audio clips.
Sound is made together with the picture, and clips can run up to 30 seconds, long enough for a short scene with several actions. Wan 3.0 Prime has the same abilities and returns results faster.
Wan 2.7 Image generates pictures from text, edits a photo on instruction and combines several input images into one scene.
The standard version is the faster one, and Image Pro is the higher tier of the same model. Both can return several images in one run, and both have a thinking mode for complex scenes described in words.
The two work well together. Make a character, a product shot or a setting on Wan 2.7 Image, then use it as the first frame or as a reference image for a Wan 3.0 clip.
Every version is included in Plus and above. Video costs credits per second of the clip and images cost credits per picture; the table below compares resolution, length, references, starting price and plan from the live model list.
Every Wan version compared
| Version | Resolutions | Length | Reference images | From | Plan |
|---|---|---|---|---|---|
| 480P, 720P and 1080P | 2–30 s | 10 | 10 credits / s | Plus |
Which Wan version should you use?
For video with sound: Wan 3.0
Choose Wan 3.0 for clips from text, from a first frame or a first and last frame, or from image and audio references, with sound made alongside the picture.
Both versions have the same abilities: Prime returns results faster, and the price per second differs, so compare the two rates on the Wan 3.0 page.
For long single takes: Wan 3.0
Wan 3.0 clips can run up to 30 seconds, so one take can hold a whole scene.
Write the actions as a sequence with rough timings, test at a short length and a low resolution, then generate the full clip.
For pictures and first frames: Wan 2.7 Image
Use Wan 2.7 Image for stills, edits and scenes combined from several photos.
Pick Image Pro for final images, and the standard version for speed and quick variations.
Wan versions on Zero Kit
Wan 3.0
Cinematic long-form narrative, featuring native 30-second single-shot sequences, simultaneous audio-video output, and full-modal anything-to-video generation.
Open Wan 3.0
Wan FAQ
Partly, through Alibaba. The official Wan site at wan.video has a free plan with daily check-in credits, while 1080p, longer clips and watermark removal come with its paid plans.
On Zero Kit it is included in Plus and above, and each run costs credits. Signing up is free, so you can open the editor, write a prompt and see the price before you upgrade.
Wan 3.0 makes video clips with sound, from text, frames or references. Wan 2.7 Image makes and edits still pictures.
They are separate models in the same family, and a picture from Wan 2.7 Image can become the first frame or a reference for a Wan 3.0 clip.
Yes. Wan 3.0 can start from an image you upload as the first frame, move between a first and a last frame, or use your images as references for the people, products and places in the clip.
Clips run up to 30 seconds. The image to video AI page starts the same way: upload a photo, describe the motion and pick a video model.
Use Alibaba's official Wan site at wan.video, or a platform that runs Wan for you.
On Zero Kit, on Plus and above, you create an account, pick a Wan version in the editor and generate in your browser. No API key, code or Chinese phone number is needed.
Use Wan 3.0 for long single takes and clips built from image and audio references. For other jobs:
- Seedance 2.0 from ByteDance for camera direction.
- Kling 3.0 for multi-shot clips.
- Google's Veo 3.1 for dialogue and sound effects made with the picture.
Running one prompt on each is the quickest way to compare them.
Yes, on a paid plan. Images and videos made on a paid plan can be used commercially, including product pages, ads and social posts, subject to our Terms. Output from the free plan is for personal, non-commercial use only.
No. We do not use your prompts, uploads or results to train AI models. To run a generation, the prompt and any reference files are sent to the provider of the model you picked, which processes that request under its own terms.
Last updated: