Wan 3.0 multimodal reference control
A reference brief can include up to 10 images, 5 video clips and 5 audio clips, plus a supported document or public page. Give each input a clear purpose.
Understands your media, documents, and web pages, giving video generation richer creative context.
Wan 3.0 AI video generator
Wan 3.0 is Alibaba's video-generation model. This workspace offers Text, Frames and References, so you can begin with a written scene, an opening composition or materials that guide the subject and action.
A reference brief can include up to 10 images, 5 video clips and 5 audio clips, plus a supported document or public page. Give each input a clear purpose.
The resolution choices are 480P, 720P and 1080P. Select Adaptive or a fixed 16:9, 4:3, 1:1, 3:4 or 9:16 canvas for the scene.
Use up to 20,000 characters for a detailed brief. Frame inputs, audio and a reusable seed provide additional ways to direct and compare takes.
Use product images for appearance and a document for verified facts. Build the brief around one reveal or demonstration that fits the chosen duration.
Let an audio reference define rhythm while images or video establish the performer and setting. Explain how the visible action relates to the sound.
Plan one story beat with a clear subject, action and ending. Assign references to the character, location or camera direction.
Supply a supported file or public page as context, then explain the visual story you want to draw from it. Verify facts before publishing.
Establish the subject's appearance, the setting or the visual treatment.
Show the movement, pacing or camera direction the new take should follow.
Give the scene a voice, musical rhythm or atmospheric reference.
Provide a script, product brief or factual background for the scene.
Use an accessible public page to supply context for the requested video.
Select Frames for opening and ending images, or References for media, a document or a public page.
Use Image1, Video1 and Audio1 to identify inputs, and state what each contributes to the new scene.
Choose the generation mode, length, resolution, ratio and sound, then check the credits before submitting.
Write the event the viewer should understand by the end of the clip. Establish the subject and setting, then explain the action and one camera move. A short reveal needs less time than a scene with an approach, interaction and departure. After generating, check whether that progression actually happened before adding more characters or visual detail.
Frames mode begins with a required opening image and can also use an ending image. The prompt should explain the action between those compositions, not merely describe the pictures again. For example, a camera could travel from a doorway toward a display, finishing on the detail shown in the last frame. Use References instead when the images are general guidance for identity or style rather than endpoints.
Treat the uploaded assets as parts of a brief. Identify Image1, Video1 or Audio1 and state the information each supplies: a product shape, a walking rhythm or a sound direction. Remove conflicting references and keep the subject's defining features compatible across the set. This makes the generated take easier to judge against the intended scene.
| Model | Wan 3.0 | Wan 2.7 | Seedance 2.5 |
|---|---|---|---|
| Input parameters | Text · Images · Videos · Audio · Documents / Web link | Text · Images · Videos · Audio | Text · Images · Videos · Audio |
| Output format | MP4 | MP4 | MP4 |
| Duration (seconds) | 2–30 / Auto | 2–15; References: 2–10 | 4–30 / Auto |
| Native resolution | 480P / 720P / 1080P | 720P / 1080P | 480P / 720P / 1080P |
| Aspect ratio | Adaptive / 16:9 / 4:3 / 1:1 / 3:4 / 9:16 | 16:9 / 9:16 / 1:1 / 4:3 / 3:4 | 1:1 / 21:9 / 4:3 / 3:4 / 16:9 / 9:16 / Adaptive |
| Prompt limit (characters) | 20000 | 5000 | 10000 |
These sample briefs show how to connect an asset, an action and an ending. Replace the references with your own materials and adjust the length to the movement being requested.
Use Image1 for the red enamel kettle and Image2 for the kitchen. Open on the kettle beside an empty cup, soft window light from the left. Move slowly toward the spout as a thin trail of steam rises. Keep the handle shape and red finish unchanged. End with the kettle and cup fully visible, leaving open space above for a later caption. Quiet room tone, no speech.
Use Image1 for the cyclist and Image2 for the riverside path. The cyclist in a gray jacket walks beside a green bicycle toward a wooden bench. Track from the side at walking pace in late-afternoon light. Keep the bicycle frame and jacket consistent. End when the cyclist stops beside the bench. Include distant water and soft footsteps, without music or dialogue.
Use the uploaded brief for the facts about the folding desk organizer. Show the closed organizer on a clear table, a hand opening its compartments and a final view with pens and cards arranged inside. Use a fixed camera and soft overhead light. Preserve the product's shape from the supplied images. Do not add performance claims or specifications absent from the brief; leave exact labels for editing.