Wan 3.0 video generator
- Type
- Video model
- Made by
- Alibaba
- Input
- Text, plus reference images and sounds, or a first and last frame
- Output
- MP4 video, with or without sound
- Options
- 5 aspect ratios · Duration 2–30 s · Resolution: 480p, 720p, 1080p · Sound · First & last frame
- Price
- From 250 credits
- Plan
- Paid plans
Wan 3.0 is Alibaba's video model for long takes. It renders up to 30 seconds of video in one continuous shot, at 480p, 720p or 1080p, with optional sound. Start from text, from reference images and sounds, or from a first and last frame.
What you can do with it
- Text to video. Describe a shot and get anything from a 2-second loop to a 30-second take.
- References. Attach images and sound clips of people, products, places or voices for the clip to work from. Up to 10 files per run.
- First and last frame. Switch on First & last frame, add the first frame (required) and an optional last frame, and the clip moves from one to the other.
- Sound. On by default, or switch it off.
How to use it
- Write the shot in the composer.
- Attach references, or switch on First & last frame and add your frames.
- Set aspect ratio, duration, resolution and sound.
- Check the price in the composer, then generate.
Writing prompts for Wan 3.0
A 30-second continuous shot needs a little structure.
- Describe how the shot develops. Say what happens at the start, in the middle and at the end, all in one place and one take.
- Plan the camera path. "Starts on the front door, follows her down the hallway, ends looking out the back window."
- Keep the cast small. One or two subjects are easier to keep track of over a long take.
- Write the sound. Room tone, footsteps, rain, music from a distant radio.
- Short prompt? Try Prompt expansion in the advanced settings panel, which expands a brief prompt before generating.
Example prompts:
- A slow walk through a night market: the camera starts on a steaming dumpling stall, drifts past hanging lanterns and a fruit seller, and ends on a busker playing violin. Sizzling pans, chatter, violin.
- Time passes on a quiet beach in one locked-off shot, from bright afternoon to sunset, the tide slowly rising. Waves and wind only.
Choosing settings
- Aspect ratio: 16:9 (default), 9:16, 4:3, 3:4 or 1:1.
- Duration: 2 to 30 seconds, default 8.
- Resolution: 480p, 720p (default) or 1080p.
- Sound: on by default.
- First & last frame: off by default.
- Prompt expansion (off by default) and Seed: in the advanced settings panel. Seed lets you pin the random starting point of a run.
Pricing is per second of video: 125 credits at 480p, 250 at 720p and 500 at 1080p, the same for every input mode. Multiply by the duration: a 2-second 480p clip is 250 credits, 8 seconds at 720p is 2,000, and 30 seconds at 1080p is 15,000. For long takes, it pays to test the idea at 480p and a shorter duration first. The composer shows the exact price before you generate.
For shorter clips with a choice of speed tiers, see Seedance 2.0.