Prompt guide
Wan 3.0 prompt guide: the formula, sound lines and 10 prompts
Wan 3.0 is Alibaba's newest video model on Viralance: 5, 10 or 15-second clips at 480p, 720p or 1080p with native sound, from 4 credits for a 5-second 480p test (about $1.20). Alibaba's own guide gives it a formula, entity plus scene plus motion, and writing in that order keeps a prompt from leaving gaps for the model to fill. It has no negative prompt field, so anything you do not want goes in the prompt as a plain sentence such as "No background music."
Updated November 7, 2026. Viralance is an independent platform that runs these models; it is not affiliated with their makers.
Versions this guide covers
| Version | Length | Resolution | Sound | Start + end frame | Credits |
|---|---|---|---|---|---|
| Wan 3.0 | 5 / 10 / 15s | 480p / 720p / 1080p | Yes | Yes | 4–38 |
The prompt structure
Entity + scene + motion, then aesthetics + style, then sound
- Entity
- The subject with two or three visible traits: "an elderly fisherman in a yellow raincoat, grey beard".
- Scene
- Where and when: "on a small wooden boat in a harbour at dawn, light fog".
- Motion
- What the entity does and how fast: "pulls a wet net over the side hand over hand, slowly".
- Aesthetic control
- Light, lens and shot size, in Alibaba's terms: "side light, close-up, shallow depth of field, eye-level".
- Stylization
- One style label: "photoreal film still", "claymation", "ink wash painting".
- Sound
- Voice (line, emotion, tone, speed) and effects or music, or an explicit "No dialogue. No background music."
An elderly fisherman in a yellow raincoat with a grey beard, on a small wooden boat in a harbour at dawn with light fog. He pulls a wet net over the side hand over hand, slowly, and water drips from the mesh. Side light, medium close-up, eye-level, shallow depth of field. Photoreal film still. Gulls call, water laps against the hull, rope creaks. No dialogue. No background music.
Alibaba's formula, in the order Wan reads it
Alibaba's Model Studio guide describes a basic formula (entity + scene + motion) and an advanced one that adds aesthetic control and stylization. For image-to-video the formula shrinks to motion + camera movement, because the photo already supplies entity and scene. Write the parts in that order, one or two sentences each, and the prompt stays readable for both you and the model.
Viralance leaves fal's prompt expansion for Wan 3.0 at its default, which is on, so Wan may add detail to a very short prompt. A prompt that already covers all five parts leaves less for expansion to guess.
Sound lines: say it, or say "no"
Wan 3.0 generates sound with the clip, and the studio's Sound toggle decides whether the audio flag is sent. Alibaba describes a voice as the character's line plus emotion, tone, speed, timbre and accent, and warns that if you describe no lines the model may add some on its own. That is the most common surprise: a nature clip with a voice-over nobody wrote.
- For a spoken line: who speaks, the line in quotes, then how: "she says 'We open at nine' in a warm, unhurried voice".
- For ambience only: list the sounds and end with "No dialogue." as its own sentence.
- For your own soundtrack later: end with "No background music." or switch Sound off, which also removes ambience.
- There is no negative prompt field on the endpoint we use, so exclusions only work as plain sentences inside the prompt.
Longer takes and shot lists
Wan 3.0 runs up to 15 seconds (19 credits at 720p, 38 at 1080p). Alibaba documents a multi-shot notation, "Shot 1 [0–3 s] …, Shot 2 [3–7 s] …", for Wan's newer models. It can produce cuts inside one clip, but timing is approximate and results vary; for exact edits, generate separate clips and cut them yourself.
Price climbs fast with resolution: 5 seconds is 4 credits at 480p, 7 at 720p and 13 at 1080p. Draft at 480p, then re-run the keeper at the size you need.
Photos keep their shape
In image-to-video Viralance sends your photo as the start image with an adaptive aspect ratio, so a vertical photo stays vertical and a landscape photo stays landscape; crop before uploading if you need 9:16. A 5-second 720p photo clip is 7 credits. Describe the motion and the camera, not the scene: "the curtains sway, the camera drifts slowly toward the window".
We do not yet have enough Wan 3.0 jobs in our generation data to publish a success rate or wait time, so test on a short, cheap setting first.
Camera words Wan 3.0 follows
| Write | Use it for |
|---|---|
| Eye-level / low angle / high angle | Alibaba lists shot angle under aesthetic control; low angle gives products and characters weight. |
| Close-up / medium shot / wide shot | Name the shot size once; Wan keeps it steady through 5 seconds. |
| Camera pushes in | Image-to-video reveals and listing photos; pair with "slowly". |
| Camera pulls back | Ending a 10 or 15-second take on a wide view of the place. |
| Tracking shot | Following a walking or running subject; say left-to-right or toward the camera. |
| Orbit around the subject | Product turntables; works best on a simple background. |
| Static camera | Dialogue and close physical detail, so the motion stays in the subject. |
| Handheld | Documentary or vlog feel; keep it to one subject in frame. |
| Aerial / drone shot | Landscapes and coastlines in text-to-video. |
Start and end frames
Wan 3.0 takes an optional end frame in Image to video mode, priced like an image-to-video clip (7 credits for 5 seconds at 720p). Upload the first frame, turn on the End frame slot, add the last frame, and write what happens between them using the motion part of the formula.
Because Wan also generates sound, describe the sound of the change: the creak of a door, a whoosh as a scene fills with light. Otherwise the transition can play over unrelated ambience.
- Both frames must share framing, lens and lighting; only the thing that changes should differ.
- Match the aspect ratio of the two images, since Wan keeps the start photo's shape.
- 10 seconds gives a smooth change; 5 seconds suits a quick reveal.
- Only Kling 4.0 takes keyframes between the two ends. With Wan, split a three-stage change into two clips.
Refused or misread prompts, and the fix
Voices appeared that you did not write
Wan adds lines when none are described. End the prompt with "No dialogue." or turn Sound off in the studio.
Music you did not want
Add "No background music." as its own sentence. Wan 3.0 has no negative prompt field on our endpoint, so exclusions must be in the prompt.
The provider refused the prompt or photo
Named people, real faces without permission, minors, sexual content and other brands are the usual causes; credits for provider refusals are refunded. See why AI video prompts get rejected.
Shots in a "Shot 1 / Shot 2" prompt landed at the wrong time
Timing in multi-shot prompts is approximate. Use fewer shots, give each a clear action, or generate them as separate clips.
The style drifted halfway through
Put one style label in the prompt and repeat it in the last sentence; on 15-second takes, avoid mixing styles ("anime" and "photoreal").
10 Wan 3.0 prompts to copy
Each one opens in the studio with the model, mode and prompt filled in. Photo prompts need your own image uploaded first.
- Wan 3.0 · text · 5s · 720p · 7 credits
1. Harbour fisherman, ambience only
An elderly fisherman in a yellow raincoat with a grey beard, on a small wooden boat in a harbour at dawn with light fog. He pulls a wet net over the side hand over hand, slowly. Side light, medium close-up, eye-level. Photoreal film still. Gulls call, water laps against the hull. No dialogue. No background music.
The full formula, with sound and exclusions.
Use this prompt - Wan 3.0 · text · 5s · 720p · 7 credits
2. Bakery owner greeting
A baker in a flour-dusted apron, mid-forties, behind the counter of a small bakery in the early morning, loaves on wooden shelves. She looks at the camera and says "First batch is out, come early." in a warm, unhurried voice. Static camera, medium shot, warm window light. Oven hum, a bell over the door.
A voice described by line, emotion and speed.
Use this prompt - Wan 3.0 · text · 5s · 1080p · 13 credits
3. Ramen close-up
A steaming bowl of ramen on a dark wooden counter, chopsticks lift a tangle of noodles and broth drips back into the bowl. Steam curls into the light. Close-up, low angle, shallow depth of field. Photoreal food cinematography. Broth dripping, soft restaurant murmur. No dialogue. No background music.
1080p for a menu hero shot.
Use this prompt - Wan 3.0 · photo · 5s · 720p · 7 credits
4. Listing photo drift
The camera pushes in slowly toward the fireplace. The room, furniture, windows and light stay exactly as in the photo. No people, no new objects. Quiet room tone. No dialogue. No background music.
Image-to-video formula: motion plus camera only.
Use this prompt - Wan 3.0 · text · 10s · 720p · 13 credits
5. Claymation fox
A small claymation fox with a knitted scarf walks through a paper-craft autumn forest, leaves falling around it, and stops to sniff a mushroom. Tracking shot from the side, eye-level. Claymation, stop-motion texture, warm light. Soft footsteps on leaves, gentle wind. No dialogue.
One style label carried through 10 seconds.
Use this prompt - Wan 3.0 · text · 10s · 720p · 13 credits
6. Two-shot story in one clip
Shot 1 [0–5 s]: a cyclist in a blue jacket rides along a coastal road at sunrise, wide shot from a drone. Shot 2 [5–10 s]: close-up of the cyclist smiling as the sun comes over the sea. Photoreal. Wind and tyres on asphalt. No dialogue. No background music.
Alibaba's multi-shot notation; cut timing is approximate.
Use this prompt - Wan 3.0 · text · 10s · 480p · 7 credits
7. Ink wash landscape
Misty mountains and a lone pine tree on a cliff, a crane flies across the valley and clouds drift slowly. Wide shot, camera pans slowly to the right. Traditional Chinese ink wash painting style, rice paper texture. Soft wind, one distant bird call. No dialogue.
Stylized draft at 480p before a larger run.
Use this prompt - Wan 3.0 · text · 5s · 720p · 7 credits
8. Phone-in-hand product review
Handheld vertical video. A young man in a grey hoodie sits at a desk and turns a matte white water bottle in his hand, then taps the lid. He says "Twelve hours later, still cold." casually, slightly amused. Natural window light. Room tone. No background music.
UGC style with a fictional creator and unbranded product.
Use this prompt - Wan 3.0 · photo · 10s · 720p · 13 credits
9. Day to night, start and end frame
The street goes from late afternoon to night: the sky darkens, shop signs and street lamps switch on one by one, and the wet pavement starts reflecting the lights. Static camera. Traffic hum grows quieter. No dialogue. No background music.
Upload a day photo as the start frame and a night photo from the same spot as the end frame.
Use this prompt - Wan 3.0 · text · 15s · 1080p · 38 credits
10. Desert storm, 15 seconds
A lone camel caravan crosses tall red dunes as a sandstorm wall rises on the horizon and slowly swallows the sun. Aerial shot, the camera pulls back over the whole take to show the scale. Photoreal, hazy orange light. Howling wind, sand hiss, camel bells. No dialogue. No background music.
The longest, sharpest setting: 38 credits, so test at 480p first.
Use this prompt
Running a Wan 3.0 prompt on Viralance
- 1
Write entity, scene and motion
Describe the subject with a few visible traits, where and when it is, and what it does, in that order.
- 2
Add aesthetics and one style
Add shot size, angle and light, then a single style label such as photoreal or claymation.
- 3
Write the sound or exclude it
Describe lines with emotion and pace, list effects, or end with "No dialogue." and "No background music." Wan has no negative prompt field.
- 4
Draft at 480p
Run 5 seconds at 480p for 4 credits, fix what is off, then re-run at 720p or 1080p.
Wan 3.0 prompt questions
What is the best prompt formula for Wan 3.0?
Alibaba recommends entity plus scene plus motion, with aesthetic control and stylization for more control, and a sound description when you want audio. For image-to-video, describe only motion and camera movement.
Does Wan 3.0 support negative prompts?
Not on the endpoint Viralance uses. Write exclusions as sentences in the prompt, such as "No dialogue." or "No background music."
Why does my Wan 3.0 clip have a voice I did not write?
Wan can add lines when the prompt describes none. End the prompt with "No dialogue." or switch Sound off.
How long and how sharp can Wan 3.0 videos be?
5, 10 or 15 seconds at 480p, 720p or 1080p, with native sound.
Can Wan 3.0 use a start and end frame?
Yes. In Image to video mode you can add an end frame at the same price as an image-to-video clip.