Skip to main content
Model family4

WAN

Long takes from a prompt, a first frame or references for picture and sound

WAN is Alibaba Tongyi Lab’s model line. WAN 3.0 and its Prime tier render long clips with a soundtrack, from a written prompt, a first and last frame, or references — images, clips and audio. WAN 2.5 is kept as a silent row that can take an audio file as an input.

Alibaba · Tongyi Lab
Strengths and limits

Strengths and limits

Key Features

  • Long single takes: WAN 3.0 and Prime render clips well past the length most families here offer.
  • Starts from whatever you have: a prompt, a first and last frame, reference images, reference clips or reference audio.
  • Sound is generated with the picture on 3.0 and Prime, and can be switched off.

Where it falls down

  • WAN 2.5 renders no sound of its own; it can take an audio file you supply instead.
  • Reference clips and reference audio each have a total length limit, listed on the release page.
  • The video rows top out at 1080p — there is no 4K option.
Model6

Model

Capability tables and prices are read from the live model catalogue.

WAN — Model family
ModelToolsDurationResolutionAudioCredits
WAN 2.7 Image ProReference-To-Image—2Kfrom 2 credits · ≈ $0.11
Wan 3.0Text-To-Video4–30s480p · 720p · 1080pfrom 5 credits · ≈ $0.29
Wan 3.0 PrimeText-To-Video4–30s480p · 720p · 1080pfrom 7 credits · ≈ $0.40
Wan 3.0Image-To-Video4–30s480p · 720p · 1080pfrom 5 credits · ≈ $0.29
Wan 3.0 PrimeImage-To-Video4–30s480p · 720p · 1080pfrom 7 credits · ≈ $0.40
WAN 2.5Image-To-Video5–10s480p · 720p · 1080pfrom 5 credits · ≈ $0.29
Tools3

Tools

Start from text

Frequently Asked Questions

WAN 3.0 for most work: it renders long clips with sound and takes references. WAN 3.0 Prime is a second tier of the same release at a higher price, and WAN 2.5 is a silent row that can take an audio file you supply.

Ready to Get Started?

Every tool shows its price in credits before you run it.