Qwen Image 3

Wan 3.0 text-to-video API · Live

Wan 3.0 Preview guide with an available video generator

Create videos directly with Alibaba Wan 3.0 through fal. The workspace includes camera, pacing, aspect ratio, duration, synchronized audio, account credits and playable results.

Public preview

Watch the Wan 3.0 preview

The clip shows a fast-paced professional kitchen scene with dialogue, ambient sound and continuous camera movement.

This embedded clip came from an attributed community preview before the official listing. The Wan 3.0 launch is now independently confirmed by Alibaba Cloud, but this clip was not generated by this site.

View original source post

What Alibaba Cloud officially confirms

The official model update record now confirms the launch date, model ID, unified workflows and maximum duration without relying on community inference.

01

Official model ID and launch date

Alibaba Cloud Model Studio added wan3.0-video in China (Beijing) on August 6, 2026.

02

One All-in-One video model

The same model supports text-to-video, image-to-video from a first frame or first and last frames, and reference-to-video workflows.

03

Preview output up to 30 seconds

The official listing confirms a maximum generation duration of 30 seconds. Preview availability, regional scope and evolving parameters should be checked before production use.

Production comparison

Wan 3.0 Preview versus Wan 2.7

Wan 3.0 is the new Beijing Preview model. Wan 2.7 remains the documented production baseline for teams that need established task-specific APIs, regions and output parameters today.

Open Alibaba Cloud's current Wan model catalog
01

Text to video

wan2.7-t2v supports 720p or 1080p output, 2–15 second duration, 30 fps, multi-shot narrative and synchronized audio.

02

Image to video

wan2.7-i2v supports first-frame generation, first-and-last-frame control and video continuation, with multimodal image, audio and video input.

03

Reference to video

wan2.7-r2v supports multi-entity references, appearance and voice continuity, 720p or 1080p output and 2–10 second clips.

04

Video editing

wan2.7-videoedit supports instruction-based editing and style transfer for 2–10 second videos at 720p or 1080p.

Wan 3.0 generator

Create a video with Wan 3.0

Describe the action, choose one camera language and set the pacing, format and duration. New tasks call the native Wan 3.0 endpoint; failed provider tasks restore consumed credits.

Wan 3.0 video scene composer

Delivery format
Duration

150 credits for 5 seconds · 720p · failed provider tasks are refunded

Your scene is ready to direct

Write what changes over time. The completed video will appear and play here.

Sources and verification boundary

The Alibaba Cloud model update is the primary evidence for launch status. The general video catalog provides the established Wan 2.7 comparison, while the community post is retained only as the embedded demo source.

Wan 3.0 FAQ

Has Wan 3.0 officially launched?

Yes. Alibaba Cloud Model Studio listed wan3.0-video in Preview for China (Beijing) on August 6, 2026.

Can the Wan 3.0 API be called now?

Yes. This site submits new tasks to alibaba/wan-3.0/text-to-video on fal and keeps each task associated with that exact model ID.

Which workflows does Wan 3.0 support?

The official listing describes one All-in-One model for text-to-video, first-frame image-to-video, first-and-last-frame image-to-video and reference-to-video generation.

Does Wan 3.0 support 30-second video, 1080p and native audio?

Alibaba Cloud explicitly confirms generation up to 30 seconds. Resolution and audio parameters should be taken from the live Preview API documentation or console rather than inferred from the community demo.

Is the video on this page generated by this website?

No. It is a locally hosted copy of an attributed community preview, displayed as a demo source alongside the independently verified official launch record.

How does the Wan 3.0 generator work on this website?

The generator submits a Wan 3.0 task, deducts the displayed site credits, polls for completion and plays the returned video. Provider failures mark the task as failed and restore the consumed credits.