The Formula
No official formula exists. A working order: Subject and action, then Environment, then one Camera move, then Sound and Dialogue.
- Subject and action
- Who moves and what they do over the clip, written as motion over time.
- Environment
- Place, light and the few details that matter to the shot.
- Camera move
- One specific move and what it ends on.
- Sound
- Concrete sources tied to visible actions, such as rain ticking against glass.
- Dialogue
- The speaker named and the exact words in quotation marks.
The Rules
There is no official syntax
Alibaba has published no prompt guide and no tag or bracket syntax for Wan 3.0. Write natural cinematic prose.
Lead with motion over time
Open with the subject and what they do across the clip, then the environment. A static description gives the model nothing to animate.
Use one camera move with a visible result
Name one move and where it ends, such as a slow push in that settles close on her hands.
Name sound sources, not moods
Sound is generated with the video when you turn it on. Name what makes the sound, tied to what is on screen.
Quote the line and name the speaker
A visible speaker is lip-synced. Mark off-screen lines as voiceover.
Stage long clips
Wan 3.0 runs up to 30 seconds. Past about 10 seconds, write ordered stages, each with one main change and a clear end state.
Skip the padding
The model expands prompts by default, so extra adjectives add little. State what matters and stop.
Phrase everything positively
Wan 3.0 takes no negative prompt, so describe what you want to see rather than listing what to avoid.
A Template You Can Copy
[Subject] [action as motion over time] in [environment]. [Light and the details that matter]. The camera [one move] until [visible result]. [Speaker] says, [delivery], "[exact words]". [Concrete sound sources tied to visible actions].
A Real Example
The shot: Night, rain-streaked diner window. A woman in a red coat sits alone in the last booth, turning a ring on the table. Slow push in from the window. Mara says: "You kept it." (6 seconds, 16:9)
A woman in a red coat sits alone in the last booth of a rain-streaked diner at night, slowly turning a ring between her fingers on the table, her eyes lowered in quiet thought. Outside the window, rain runs down the glass in thin streaks, blurring the soft glow of a distant neon sign; warm diner light spills over her face and hands. The camera performs a slow push in from outside the window, gliding through the rain-streaked glass until it settles close on her hands and face. As it nears, she lifts her eyes and says, quietly, "You kept it." Rain ticks steadily against the glass, the diner's ambient hum and a distant clink of cups fill the quiet background.
It leads with her action, sets the environment and light, then gives one camera move with a clear end point (settling close on her hands and face). The line is quoted with the speaker and delivery in plain prose, and the sound is named by source: rain on the glass, the diner hum, the clink of cups.
Dialogue
Wan 3.0 has no dialogue syntax of its own, so use plain labeled speech: name the speaker, add the delivery and put the exact words in quotation marks. A visible speaker is lip-synced; call off-screen lines voiceover. Wan 2.7 renders without sound, so leave lines out for it.
She says, quietly, "He's not coming."
Camera Moves
| Move | How to write it for Wan 3.0 |
|---|---|
| Slow push in | The camera performs a slow push in from outside the window until it settles close on her hands and face. |
| Pull back | The camera pulls back until the whole empty platform is in frame. |
| Tracking | The camera tracks alongside the cyclist until she stops at the light. |
| Pan | The camera pans left to reveal the open doorway behind him. |
| Static | The camera holds still as the bus pulls away and leaves her alone at the stop. |
Common Mistakes
| Mistake | Fix |
|---|---|
| Padding the prompt with adjectives. | The model expands prompts itself; keep the action, one camera move and the sound. |
| Writing a negative prompt such as "no blur, no extra people". | Wan 3.0 takes no negative prompt. Describe what should be in the frame instead. |
| Giving Wan 2.7 a spoken line. | Wan 2.7 renders without sound. Drop the line or use Wan 3.0. |
| Abstract sound like "tense atmosphere". | Name the sources: "rain ticks against the glass, a fridge hums". |
| One long clip with no structure. | Past about 10 seconds, write ordered stages with one change each. |
Length, Resolution and Price
Wan 3.0 makes 2 to 30 second clips with optional sound and lip-synced speech, accepts an end frame and up to 10 references, and costs $0.05 per second at 480p, $0.10 at 720p or $0.20 at 1080p, in 16:9, 4:3, 1:1, 3:4 or 9:16. Wan 3.0 Prime costs $0.068, $0.14 or $0.28 per second, about 40% more, with no published quality claim for the difference. Wan 2.7 runs 2 to 15 seconds without sound at $0.10 per second at 720p or $0.15 at 1080p (checked 2026-10-22). Every model side by side is on the model spec sheet.