Native 30-second storytelling
Generate an integer duration from 2 to 30 seconds, or use smart duration so the model selects a suitable length.
Verified Aug 27, 2026
Wan 3.0
A practical hub for Alibaba's all-in-one video model: understand the 30-second multimodal workflow, build requests, estimate cost, write stronger prompts, and turn documents into production-ready plans.
30smaximum single generation
1080Poutput resolution
6input types
A/Vnative audio-video
InputText · image · video · audio · file · link
One modelwan3.0-video
OutputUp to 30s · 1080P · native audio
One model for the complete video request
Wan 3.0 replaces a menu of task-specific endpoints with one all-in-one model. The media type and the intent in your prompt route text-to-video, first-frame animation, first-and-last-frame control, multimodal reference, document understanding, editing, or extension.
Wan 3.0 is a hosted generation model, not a free public iframe.The official access path is Alibaba Cloud Model Studio. A dependable anonymous Wan 3.0 Space was not available at verification time, so this site focuses on useful zero-cost planning tools instead of presenting an unstable demo as permanent free generation.
What changed in Wan 3.0
The release expands both the creative canvas and the amount of source material a single request can understand.
Omni-reference input
Combine up to 10 images, five video clips, and five audio clips while addressing each asset by type and order.
Documents and web pages
Use a file or a public web page as structured source material for explainers, ads, presentations, and narrative video.
Consistent people and objects
Reference guidance targets faces, props, spaces, style, and relationships across a longer sequence.
Native synchronized sound
Dialogue, effects, ambience, and music can be generated with the picture; disabling audio does not change the price.
Editing and continuation
Modify visuals, plot, dialogue, or extend a supplied video through prompt intent without choosing a separate model family.
The all-in-one task map
The same model ID routes work from the media types supplied in input.media and from clear action words in the prompt.
| Task | How it is triggered | Best practice |
|---|---|---|
| Text to video | Prompt only | Set ratio, resolution, duration, and audio directly. |
| First frame | first_frame | Use adaptive ratio to preserve the image composition. |
| First and last frame | first_frame + last_frame | Describe the transition and motion between the two anchors. |
| Omni reference | reference_image / video / audio | Refer to Image 1, Video 1, and Audio 1 by their type-specific order. |
| File or web page | file or link | One public file or page per request; file and link are mutually exclusive. |
| Edit or extend video | reference_video + edit intent | Use adaptive ratio; smart duration is useful for preserving or extending timing. |
Use the Wan 3.0 toolkit
Each page solves a different planning decision without consuming model quota.
API Builder Build a valid asynchronous cURL and JSON request without exposing an API key. Pricing Calculate Standard, Prime, promotion, input-video, output-video, and free-quota costs. Prompt Guide Assemble motion, camera, timing, reference, dialogue, and sound into a copyable prompt. vs Wan 2.7 Compare the official Wan 3.0 and Wan 2.7 limits, APIs, and text-to-video costs. Document to Video Validate a document locally, extract useful structure, and create a document-to-video brief.
What still needs caution
The launch is substantial, but the operational boundary matters as much as the headline features.
Official generation requires a regional Model Studio workspace and API key; the resource pages here do not proxy paid requests.
The 30-second free quota applies only in the Beijing region, combines input and output video seconds, and expires after 90 days under the official rules.
Alibaba notes that audio texture and on-screen text accuracy are still improving, so review dialogue, labels, charts, and typography before delivery.
Generated result URLs and task IDs are available for 24 hours; production systems should download outputs to durable storage.
Wan 3.0 questions
Is Wan 3.0 free?
The model is pay-as-you-go. Beijing accounts may receive a one-time 30-second quota valid for 90 days; the international region currently has no official free quota. These planning tools are free and make no model call.
Can Wan 3.0 generate a full 30-second clip?
Yes. Without input video, duration accepts any integer from 2 to 30. With input video, the combined input-video and output-video duration cannot exceed 30 seconds.
Does Wan 3.0 accept PDF and PowerPoint files?
Yes. Official formats include common office documents, PDF, text, Markdown, Keynote, Pages, and Numbers. One file up to 100 MB and up to 50 validated pages is accepted.
Does Wan 3.0 have native audio?
Yes. Audio is enabled by default. Prompts can direct dialogue, voice character, effects, ambience, and music; audio on or off does not change billing.
Official sources and verification boundary
Facts and prices on these pages are dated snapshots from Alibaba Cloud and Model Studio. Confirm the console before a paid production run.