30s
Native video window
The Public Beta preview announces clips of up to 30 seconds in one generation.
Coming soon
Wan 3.0 creation preview
Preview the next-generation Wan 3.0 creative generator
Wan is Alibaba's video-generation model family. Wan 3.0 is expected to bring major improvements to AI video creation, including higher-quality output, stronger consistency, and more advanced creative control, with full access coming soon.
Wan 3.0 is coming soon. The mode-specific workspaces are already available with other supported models.
20
Cross-format references
Omni-Creation is presented as accepting up to 20 project assets in one task.
8
Input categories
Text, images, audio, video, documents, spreadsheets, slides, and web pages.
Wan 3.0
What is Wan 3.0?
Wan 3.0 is the latest publicly tested visual-generation model and creative platform from Alibaba's Wan team. It covers video generation, image generation, and editing existing content through text instructions or reference material.
Public Beta means the product is open for general users to apply for testing, while features, pricing, eligibility, and output quality may still change. It is not the same as a final commercial release, public API, or open-source model.
Native 30-second video
Announced functionality
Up to 30 seconds in one generation with intelligent duration control
Practical value for creators
Reduces identity, lighting, and narrative jumps caused by stitching many short clips
Omni-Creation
Announced functionality
Up to 20 reference assets
Practical value for creators
Bring characters, products, scripts, presentations, and brand material into one task
Reference consistency
Announced functionality
Reproduce character and object details
Practical value for creators
Better suited to ads, character shorts, and work that depends on a stable visual identity
Precise video editing
Announced functionality
Instruction- and reference-driven changes
Practical value for creators
Modify the picture, story, or environment without restarting every generation
Together, these capabilities shift the focus from generating one attractive shot to completing content that can be controlled, reviewed, and revised.
Wan 3.0 Studio
From a single prompt to a controllable production flow
The Public Beta direction moves beyond isolated clips: bring planning material into the generation context, preserve important visual anchors, and refine an existing result instead of restarting every time.
Native 30-second creation
A longer single generation can reduce the identity, lighting, and pacing breaks caused by stitching many short clips.
Omni-Creation context
Combine creative briefs, brand documents, reference images, videos, and audio instead of compressing everything into one prompt.
Reference-led consistency
Keep characters, products, environments, and brand cues visible throughout a multi-step creative workflow.
Wan 3.0 Public Beta
A broader image and video creation system
The preview describes Wan 3.0 as a connected creation environment rather than a single text-to-video endpoint.
Precise video editing
Use instructions and references to modify the picture, story, or environment without rebuilding the entire concept.
Image generation and editing
Develop characters, storyboards, brand visuals, multi-image compositions, and selected-area edits before motion.
Production verification
Long text, formulas, figures, brand details, and document-derived facts still require human review before publishing.
Omni-Creation preview
Let project material become part of the prompt
The announced workflow can take up to 20 references across eight content categories. That creates a clearer path from a brief, storyboard, product material, and brand guide to a generated or edited result.
TextImagesAudioVideoDocumentsSpreadsheetsSlidesWeb pages
Input
4 input types
Audio track
AAC · 5.0s
30s
Native video window
20
Cross-format references
8
Input categories
Workspaces
Choose a workflow to preview
Each URL keeps its creation mode fixed. Choose another currently available model now; Wan 3.0 remains marked as coming soon in the model list.
Coming soon
Text to Image
Turn a written visual direction into a still-image brief.
PreviewComing soon
Image to Image
Prepare a reference-led image transformation workflow.
PreviewComing soon
Text to Video
Plan a cinematic shot from subject, action, camera, and atmosphere.
PreviewComing soon
Image to Video
Animate a starting frame while preserving composition and identity.
PreviewComing soon
Multi Reference
Organize visual references for consistency across a generated shot.
PreviewComing soon
Video to Video
Prepare an existing clip for guided editing and restyling.
PreviewFAQ
Wan 3.0 frequently asked questions
What Public Beta means, what has been announced, and which details are still pending.
Has Wan 3.0 officially launched?
Wan 3.0 entered Public Beta on August 6, 2026, and users can apply through the official website. Public Beta is still a testing stage, so pricing, features, and eligibility are not necessarily final.
What is the longest video Wan 3.0 can generate?
The announced native video length is up to 30 seconds with intelligent duration control. Actual selectable durations, resolutions, and mode-specific limits should follow what appears in the user's account.
Which reference material can Wan 3.0 read?
The announced categories include text, images, audio, video, documents, spreadsheets, presentations, and web pages, with up to 20 reference assets. Complex tables and exact figures still require human verification.
Is Wan 3.0 open source?
No official Wan 3.0 open-source project or model weights have been identified. Earlier Wan releases included open models, but that does not mean Wan 3.0 uses the same release model.