Wan 3.0 Image-to-Video is the image surface of Alibaba Cloud's newest Wan model, served through DashScope. It animates a still you supply and generates a soundtrack with it, holding a take for up to 30 seconds. Generation runs in the background, so the clip appears when it is ready rather than immediately.
What it's good at
Wan 3.0 starts from a frame you already have and moves it, so the look of the first frame is settled before generation begins. It suits:
- Animating artwork and stills: a photograph, render or illustration becomes a moving shot without a reshoot.
- Controlling how a shot ends: an optional last frame, so the clip travels towards a composition you choose.
- Longer takes: up to 30 seconds from one frame, with sound generated alongside.
Use Wan 3.0 Text-to-Video when you have no starting frame, or Wan 3.0 Reference-to-Video when the image is a subject to carry through a new scene rather than the opening frame.
Inputs
- First frame: the still the clip opens on. Required.
- Last frame: optional. Give one when you want the shot to arrive at a particular composition.
Settings
- Resolution: sets the frame size of the clip. Choose 480P, 720P or 1080P. Default: 480P. Each step up costs more per second.
- Duration: sets the length of the clip. Choose any whole number of seconds from 2 to 30. Default: 5. Longer clips cost more, because the rate is per second of output.
- Watermark: off by default. Turn it on to burn the vendor's mark into the lower right corner.
Practical tips
- Upload the frame at the tier you intend to render. A small source limits what the model can hold at 1080P.
- Describe the motion you want, not the picture. The frame already carries the composition.
- Use a last frame when the shot has somewhere to go. Leave it off when you want the model to choose.
- Draft at 480P and a short duration, then re-run at the tier and length you need.
