Video model by Alibaba
Wan 3.0
Alibaba's video model for 30-second clips from text or images. Generates with optional sound in the same pass, maintaining character and style across shots.
- Makes
- Text + Image to video
- Starts at
- $0.056 per second of video
- Plans
- Every plan, including Free
- Released
- August 10, 2026
Wan 3.0 is Alibaba's answer to longer, multi-shot AI video generation. It breaks the 15-second ceiling and holds character continuity across twice the length, making it possible to fit a whole product demo or a scene with setup and payoff into one clip.
Why creators choose Wan 3.0
Wan 3.0 shifts what's possible in a single generation. Thirty seconds used to mean cutting five or six clips together; now it means one cohesive piece with fewer seams.
Up to 30 seconds without breaking continuity
A single Wan 3.0 generation runs 2 to 30 seconds, which holds a complete thought. Product introductions, short dramas, and ad spots all fit in one pass. Fewer clips mean fewer transitions to hide or patch. That's why Wan 3.0 works better for brand spots and demo reels than faster models capped at 15 seconds.
Visual consistency across the full clip
Wan 3.0 maintains character faces, object details, and scene props across the entire video. Feed it a reference image and the model keeps your subject consistent from frame one to frame thirty. That's the difference between a polished spot and an obvious patchwork of different takes. On Craze, one generation supports up to 10 reference images, giving you control over product appearance, character features, and brand assets across all 30 seconds.
Optional sound in the same pass
Wan 3.0 generates audio and video together, so sound is ready the moment the clip is done. On Craze you can turn audio off if you're adding music or voiceover in the edit, or keep it on for immediate playback. This matters for client work where a voiceover is planned but sync matters, or for social content where in-video sound keeps viewers watching.
Multiple aspect ratios for any platform
The model supports 16:9, 4:3, 1:1, 3:4, and 9:16. Shoot one brief for vertical, square, and widescreen at once. For clients who need repurposing across platforms without additional generations, this saves time and keeps framing choices consistent.
How to use Wan 3.0 in Craze
You don't need to master Wan's prompt style. You direct the video, and Craze handles the model-specific details.
- Start a project and describe the video you want to make. Add reference images if you want to steer character or product appearance.
- Say you want Wan 3.0, or pick it in the Models panel. If you don't name a model, Craze shows compatible options with prices before it plans.
- Craze plans your clips, writes the prompt for each one with your references attached, and generates the full video.
- Review the clip in the editor. Regenerate any shot that misses, keep what works, and keep editing.
If a camera move or beat matters, write it into your brief: "Start wide on the room, then push in on the action" is more useful than "make it cinematic."
Wan 3.0 pricing
You pay per second of video. The cost depends on resolution; 480p is the most affordable, and 1080p costs the most. Longer clips cost proportionally more. $0.056 per second of video.
Use the cost calculator on this page to estimate the price of any clip before you generate. Craze groups multiple clips together, so if you're generating a 30-second piece as one clip or three 10-second clips, the calculator shows the difference.
Who Wan 3.0 is for
Product demos and testimonials
A 30-second clip holds an introduction, the product in action, and a closing shot. Wan 3.0 keeps logos and packaging consistent, so the same product appears throughout.
Short-form branded content
A vertical 30-second spot has room for hook, action, and call-to-action. Craze can generate for mobile, desktop, and square all at once.
Story-driven content and micro dramas
Wan 3.0's multi-shot capability lets you build scenes with camera movement and continuity. The same character stays the same across all 30 seconds when you provide a reference.
Wan 3.0 vs other models
For longer clips with multi-shot capability, Wan 3.0 and Seedance 2.5 both run 30 seconds, but Seedance 2.5 supports more reference images and videos for complex continuity work. If resolution is your priority, Seedance 2.0 goes up to 4K for clips up to 15 seconds.
When speed matters, Wan 3.0 Prime generates the same results at roughly 1.5x speed, with the same controls and quality. You pay more per second for the acceleration. Choose Wan 3.0 if you have time, and Wan 3.0 Prime if your deadline is tight.
Wan 3.0 specs on Craze
| Task | Inputs | Resolution | Length | Aspect ratios | Sound |
|---|---|---|---|---|---|
| Text to video | Text prompt | 480p, 720p, 1080p | 2 to 30 s | 16:9, 4:3, 1:1, 3:4, 9:16 | On or off |
| Image to video | First frame, Last frame | 480p, 720p, 1080p | 2 to 30 s | Follows the input | On or off |
Frequently asked questions
Wan 3.0 is a video generation model from Alibaba that creates clips up to 30 seconds from text prompts or images. It generates video with optional synchronized audio, maintaining character consistency and visual style across shots.
Alibaba released Wan 3.0 to public beta on August 10, 2026, and launched it widely on August 24, 2026.
On Craze, Wan 3.0 generates clips from 2 to 30 seconds. For longer pieces, Craze creates multiple clips and cuts them together on the timeline.
Wan 3.0 generates at 480p, 720p, or 1080p on Craze. The cost per second increases with resolution, with 480p being the most affordable option.
You pay $0.056 per second of video. The rate depends on the resolution and duration. The cost calculator on this page estimates the price before you generate.
every Craze plan, including Free. Every plan includes monthly credits for generations, and Wan 3.0 draws from them per second of video you create.
Yes. Wan 3.0 can generate audio together with the video in a single pass. Sound is optional on Craze, so you can disable it if you plan to add music or voiceover in the edit.
Yes. On Craze, Wan 3.0 supports reference images and can take a first and last frame to anchor your clip's start and end points.
More video models
Seedance 2.5
ByteDance Seed Team
- Text to video
- Image to video
- Video extension
- Reference to video
- Video editing
From $0.114 per second of video
Seedance 2.0
ByteDance Seed Team
- Text to video
- Image to video
- Video extension
- Reference to video
- Video editing
From $0.075 per second of video
Seedance 2.0 Fast
ByteDance Seed Team
- Text to video
- Image to video
- Video extension
- Reference to video
- Video editing
From $0.06 per second of video