Describe visible change
Replace vague words such as ‘dynamic’ with who moves, in which direction, how quickly, what reacts, and what the final state becomes.
Verified Aug 27, 2026
Wan 3.0 Prompt Guide
Turn a loose idea into a production-ready Chinese or English prompt. The builder organizes subject, scene, motion, camera, timeline, references, dialogue, sound, and continuity without calling an API.
20Kprompt limit
中 / ENofficial prompt languages
30ssingle-generation canvas
0API calls
InputText · image · video · audio · file · link
One modelwan3.0-video
OutputUp to 30s · 1080P · native audio
Build a structured video prompt
Fill only the fields that matter. A clear sequence of visual and audio decisions is usually more useful than a pile of unrelated style adjectives.
Start from a preset
Prompt output language
How to write for a 30-second multimodal model
Wan 3.0 can understand more context, but it still benefits from explicit priorities and a readable timeline.
Give the camera a job
Name shot size and camera behavior: locked frame, push-in, tracking, orbit, handheld follow, rack focus, or a motivated cut.
Use time beats for long clips
For 20–30 seconds, divide the idea into a small number of timed beats so the setup, interaction, payoff, and close have room.
Address references by type
Image 1 and Video 1 are both valid because images, videos, and audio are numbered separately in media-array order.
Direct sound as part of the scene
Write exact dialogue and describe emotion, pace, timbre, effects, ambience, and music timing instead of simply asking for ‘good audio’.
State continuity priorities
Name the few things that must not drift: face, clothing, product geometry, prop state, room layout, screen direction, or color palette.
Before you generate
Does every reference label match the media array order and type?
Can the described action physically fit inside the requested duration?
Are dialogue and sound cues tied to a speaker, action, or time beat?
Are the most important identity and continuity constraints explicit?
For an input image, did you focus on motion and camera instead of needlessly redesigning the frame?
For text or charts on screen, have you planned a manual review pass?
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.