Wan AI / Wan 2.5
Wan 2.5
The quality step up from 2.2, at roughly double the cost per second.
A preview-channel release with noticeably better motion coherence than 2.2. Same two modes, same prompt grammar — the difference you are paying for is that things move like they have mass.
Generation for Wan 2.5 runs in Mintdreamer today — this site is the reference, not the studio. If that changes, it will change here first.
What Wan 2.5 produces.
The prompt and parameters are the real ones we would send. The clip is empty until the render is ours — borrowing someone else's output to fill this frame would undo the only claim this site makes.
a harbour at dawn, fishing boats rocking against the pier, gulls crossing the frame, handheld
The parameters below are real. The clip isn’t generated yet — we would rather show you nothing than someone else’s output with our label on it.
The endpoints behind this version.
Model id, who serves it, and the provider's price for one unit of output — before our margin. Two entries means text-to-video and image-to-video are separate endpoints for this version.
- Model id
- alibaba/wan-2-5-t2v
- Served by
- fal
- Provider price
- $0.15 / second
- Credits
- 150 per second
WAN 2.5
- Model id
- alibaba/wan-2-5-i2v
- Served by
- fal
- Provider price
- $0.15 / second
- Credits
- 150 per second
WAN 2.5 (I2V)
How this sits against the other versions.
Wan releases are close enough together that the version number is the whole decision. This is the short version of that decision.
Wan 2.2
The version most people mean when they say Wan.
Text to video · Image to video
Wan 2.7
The current flagship — and the version search interest is climbing fastest on.
Text to video
Wan 2.6
Announced, and not something we can run for you.
Not routed here
Wan 2.7 Image
Stills from the same family, at a fraction of the cost of a clip.
Text to image