On this page
Short answer
Write a Wan 3.0 prompt in four parts: who or what is on screen, what happens, how the camera moves, and what you hear. Put spoken lines in quotation marks. Start from a picture with a first frame (or a first and a last frame), or call up to 10 saved characters by @handle, but not both in one clip.
Try this in the studio
The four-part prompt
- 1
Subject and place
Who or what is on screen, and where: "a street food cook at a night market". Add the look in plain words: light, weather, clothing.
- 2
Action
What happens, in the order it happens. For a 5 to 8 second clip, one or two actions is plenty.
- 3
Camera
One move: a slow push in, a tracking shot, a quarter orbit, a handheld close-up or a locked-off frame.
- 4
Sound
Switch sound on and say what you want to hear: a line of dialogue in quotation marks, effects, ambience.
Wan 3.0 prompts to copy
Each card shows the settings and the credit cost from the studio's own code. Change the subject, place and camera to make it yours.
Starting points
Close-up with sound
Close-up of a street food cook tossing noodles in a flaming wok at night. Sparks and steam rise, neon signs blur in the background. The camera slowly pushes in on the wok. Sound: sizzling oil, the clang of the wok, crowd chatter.
Why it works: One subject, one action, one camera move, then the sound you want to hear.
- Wan 3.0 Prime
- 5 s
- 720p
- 9:16
- 12 credits
Character line to camera
A young woman in a denim jacket sits on a park bench in golden hour light, looks into the camera and says: "I tried every app. This one actually stuck." She smiles and shrugs. Handheld medium close-up, soft background.
Why it works: Dialogue goes in quotation marks after "says", with the action around it.
- Wan 3.0 Prime
- 6 s
- 720p
- 9:16
- 15 credits
Product from your photo
The perfume bottle from the first frame stands on wet black stone. A thin beam of light sweeps across it, droplets catch the light, and the camera orbits a quarter turn to settle on the label. Quiet ambient hum.
Why it works: With a first frame, describe what moves and how; the picture already sets the look.
- Wan 3.0 Prime
- 8 s
- 1080p
- 1:1
- 39 credits
- First frame: your product photo
From one frame to another
Time passes in one continuous shot: the empty cafe in the first frame fills with people, lights come on outside, and the scene ends exactly on the last frame. Locked-off camera, natural sound of a busy cafe.
Why it works: First and last frame: say what changes in between and keep the camera steady.
- Wan 3.0 Prime
- 10 s
- 720p
- 16:9
- 24 credits
- First and last frame
Two saved characters
@mira and @leo walk side by side along a rainy city street holding coffees. @mira turns to @leo and says: "Same time tomorrow?" He laughs and nods. Tracking shot from the front, warm shop lights, light rain.
Why it works: Each @handle becomes Image 1, Image 2 in the order you name them, so say who does what.
- Wan 3.0 Prime
- 8 s
- 720p
- 16:9
- 20 credits
- Saved characters
How do you write dialogue for Wan 3.0?
Put the exact words in quotation marks after a verb like "says" or "whispers", and describe the delivery around them: "looks into the camera and says: ...". Keep a line short enough to say in the clip: about two words a second is a natural pace.
Switch sound on in the studio, or set generate_audio on the API. Sound does not change the price.
First frame, or first and last frame?
Add one picture as the first frame to animate it, or 2 pictures as the first and the last frame to move from one to the other. The picture sets the look, so the prompt only needs the motion: what moves, how the camera moves and what you hear.
How do saved characters work in a Wan 3.0 prompt?
Type @ and pick a saved character or image; you can use up to 10. Each one is sent as a reference and named Image 1, Image 2 and so on, in the order they appear, so always say what each character does.
Characters and frames are two different modes: a clip uses either saved characters or a first and last frame, not both.
Which shape and length should you pick?
Wan 3.0 renders 16:9, 9:16, 1:1, 4:3, 3:4 and 21:9: 9:16 for Reels, Shorts and TikTok, 1:1 for feeds, 16:9 or 21:9 for widescreen. Lengths are 4, 5, 6, 8, 10, 12, 15, 20, 25 and 30 seconds.
For a longer clip, write the prompt as beats: what happens at the start, the middle and the end. Draft at 720p (12 credits for 5 seconds) and render the keeper at 1080p.
What if the clip is not what you asked for?
Change one thing at a time so you can see what helped:
- Too much happening: cut the prompt to one subject, one action and one camera move.
- The camera wanders: name the move and add "steady" or "locked-off".
- The wrong character does the action: name them by @handle in the sentence where they act.
- The line is cut off: shorten it or choose a longer clip.
Frequently asked questions
How long can a Wan 3.0 prompt be?
Write what the clip needs: a few sentences covering subject, action, camera and sound. Very long prompts with many actions tend to blur together in a short clip.
Can Wan 3.0 make people talk?
Yes. Switch sound on and put the line in quotation marks after "says".
Can I use a saved character and a first frame together?
No. A clip uses either saved characters and images as references, or a first frame (and optional last frame).
How many characters can I use?
Up to 10 saved characters or images in one clip.
Does Wan 3.0 Prime need different prompts?
No. Prime is the same model in a faster mode, so the same prompt works on both.
