Wan 2.2
Wan 2.2 is Alibaba’s image-to-video model for 5 or 8 second clips at 480p or 720p, generated from a still and a prompt, with an optional second image to move toward. Use it as the first pass on any picture: it answers whether the motion you asked for is the motion you get, faster and cheaper than anything else in the studio that starts from an image.
It runs in the editor on the left at 12 credits a second for 480p and 24 for 720p, so the four possible clips are 60, 96, 120 and 192 credits. The output is a silent MP4. It does not start from text.
Why start here
Because the expensive part of image to video is finding out that the prompt was wrong, and Wan 2.2 makes that finding cost 60 credits and one minute. The framing and the motion you see at 480p are what you would get at a larger size on the same model; only the detail changes. When the 480p pass moves right, spend the larger budget on Wan 2.7 for 720p and 1080p, or on Wan 3.0 or Seedance 2.5 for a longer take.
The ending frame is the other reason. Give it a second image and the clip moves from the first to the second, which is how you get a controlled turn, a reveal, or a transition between two stills.
What it costs
Per second of output. Four settings, four prices.
| Clip | 480p | 720p |
|---|---|---|
| 5 seconds | 60 credits | 120 credits |
| 8 seconds | 96 credits | 192 credits |
Our test
The hero clip is the still from our home page animated with one sentence at 480p for 5 seconds. It cost 60 credits, came back in about 60 seconds, and is an 832×480 silent MP4 of 5.0 seconds.
The same still and sentence went through Wan 2.7, Wan 3.0 and Seedance 2.5, each with its own page, so the four can be compared on one shot. This is the fastest and cheapest of the four and the only silent one.
| Field | Value |
|---|---|
| Input | One still, 1280×720, plus the prompt in the caption |
| Settings | 480p, 5 seconds |
| Cost | 60 credits ($0.60) |
| Wall time | About 60 seconds |
| Output | 832×480, 5.0 s, silent MP4 |
Wan 2.2 or something else?
Grok Video 1.5 also animates a still, at 25 credits for 5 seconds, and adds audio; it is the one model cheaper than this for the job. Wan 2.2 keeps its place for the ending-frame move, which Grok does not offer, and for its consistency: two runs of the same still and prompt land closer together than on the draft models.
How to run Wan 2.2 here
- 1. Upload the still PNG, JPG or WebP under 10 MB, every side between 360 and 2000 pixels; the page checks this before uploading.
- 2. Add an ending frame if you want a controlled move Optional. The clip travels from the first image to the second.
- 3. Describe the motion in one sentence The model already sees the picture; tell it what moves.
- 4. Create at 5 seconds and 480p 60 credits, about a minute. Go to 720p or 8 seconds only once the motion is right.
When not to use this
- Sound. Wan 2.2 returns a silent file. Every other model here generates audio; Wan 2.7 can even follow yours.
- Text to video. It needs a still; pick Wan 2.2 and the image slot becomes required.
- Clips over 8 seconds or above 720p. Wan 2.7 goes to 15 seconds and 1080p; Wan 3.0 and Seedance 2.5 to 30 seconds.
- The take you keep, most of the time. Use it to prove the motion, then re-run the same still on a sharper model.
- A photo of somebody who has not agreed to it.
Questions people ask
How much does Wan 2.2 cost?
12 credits a second at 480p and 24 at 720p, where 1 credit is one US cent. A 5-second 480p clip is 60 credits, an 8-second 720p clip is 192. Charged when the job starts, refunded if it fails.
How long does Wan 2.2 take?
About a minute for a 5-second 480p clip on our test, which makes it the fastest image-to-video model in the studio.
Does Wan 2.2 generate audio?
No. The MP4 is silent. If the clip needs sound, generate it on Grok Video 1.5, Wan 2.7, Wan 3.0, MiniMax H3 or Seedance instead.
What is the ending frame for?
A second image the clip moves toward. Use it for a turn, a reveal, or a transition between two stills; leave it out and the model decides where the motion goes.
Can Wan 2.2 make video from text?
No. It is image-to-video only. Wan 3.0 is the Wan model that starts from a prompt.
Related tools
Checked 2026-09-23.