Wan Prompt Cheat Sheet (2026)

Keep this page open while you write Wan 2.7 prompts. It has the exact vocabulary the model understands — shot sizes, camera moves, lighting terms, and the native-audio syntax that makes Wan generate synced dialogue and sound. Wan (Tongyi Wanxiang) is Alibaba's open-weight video model, and because it ships under an Apache 2.0 license the clips are free to use commercially. Every table below is copy-paste ready, and the settings section tells you what belongs in the app rather than the prompt. For full walkthroughs, see the 35 best Wan prompts and how to prompt Wan for realistic video.

The 6-part shot formula

Wan responds best when you write like a director handing a crew a shot brief — one flowing paragraph, then the audio lines beneath it. Every strong shot breaks into six beats, in this order. The last one, style, is where the audio lines live, and native audio is what sets Wan 2.7 apart from older open models.

Stack the first five beats into one paragraph, add the audio lines underneath, and set duration, resolution, and aspect ratio in the app before you generate. Grab terms from the tables below and drop them straight in. Keep the whole prompt under about 200 words so the model does not average competing ideas. See Wan prompt templates for fill-in-the-blank skeletons built on this exact formula.

Shot sizes

Shot size controls how much of the subject and setting is visible. Pick one per shot and name it explicitly.

Camera moves — one per shot

Use exactly one move per shot. Wan follows a motion hierarchy, so stacking moves — a pan that also dollies and orbits — makes the motion unstable and shaky. Need a second move? Make it a second clip or use Wan's multi-shot mode with labeled scenes.

Lighting & lens

Pair one lighting term with one lens or style term for a consistent, filmic look. Two is plenty — more and Wan starts averaging them out.

Native audio syntax

This is Wan 2.7's headline feature: it generates dialogue, voiceover, sound effects, ambient sound, and music synced to the video. Write each layer as its own line beneath the visual paragraph. Keep dialogue short — a clip is only 5-15 seconds, so aim for one or two brief lines.

In image-to-video you can also pass a real audio file alongside the start frame, and Wan matches lip and body motion to the track instead of generating the sound itself.

Output settings & modes

These are chosen in the app or API (Alibaba Cloud Model Studio or a hosted service) before you generate — not written into the prompt text. Wan has no inline parameters like Midjourney's --ar, so typing a duration or aspect ratio in the prompt does nothing.

Example prompts

Five complete Wan shots built from the tables above. Each ends with a plain settings note in parentheses — a reminder of what to pick in the app, not text to paste into the prompt.

1. Product close-up with dialogue

A barista in a green apron sets a ceramic cup on a wooden counter and pushes it toward the camera, steam curling upward. Cozy café interior, early morning light through a tall window. Medium close-up, slow push-in. Golden hour lighting, shallow depth of field, 85mm.
Dialogue: the barista says, "One oat flat white, coming up."
SFX: espresso machine hissing, cup settling on wood.
Ambient: quiet café murmur, soft jazz.
(16:9, 1080p, 8s)

Best for: product and café ads where the spoken line sells it. Why it works: one action plus one move keeps the motion stable while Wan's synced audio does the selling.

2. Walking city scene

A woman in a red raincoat walks briskly down a rain-slicked sidewalk at night, weaving between umbrellas. Downtown street, neon signs reflecting in puddles. Wide shot, handheld follow alongside her. Low-key lighting, practical neon, 35mm, film grain.
SFX: footsteps splashing on wet pavement, a distant car horn.
Ambient: steady rain, muffled city traffic.
(9:16, 1080p, 10s)

Best for: vertical social openers and moody b-roll. Why it works: one motion, one move, and an ambient bed that grounds the whole shot.

3. Nature establishing shot

A lone hiker crosses a ridgeline as morning fog drifts through the valley below. Mountain range at sunrise. Extreme wide shot, slow crane rising to reveal the full valley. Golden hour lighting, volumetric light through the fog, 24mm.
SFX: wind gusting across the ridge, a single crow call.
Ambient: distant birdsong, a faint river below.
Music: sweeping cinematic strings.
(16:9, 1080p, 10s)

Best for: openers and travel content. Why it works: a single steady crane keeps an epic reveal coherent where a swooping move would wobble.

4. Two-person dialogue

Two friends sit across a small table, one leaning in as the other reacts. Warm bistro interior, afternoon light through a large window. Over-the-shoulder shot, static locked-off camera. High-key lighting, softbox fill, shallow depth of field, 50mm.
Dialogue: she says, "You actually did it — you quit today?" He grins, "This morning."
SFX: cutlery clinking, a chair scraping nearby.
Ambient: low bistro chatter, warm background music.
(16:9, 1080p, 8s)

Best for: character beats and skits. Why it works: two short lines fit the clip, and a locked-off camera lets Wan spend its motion budget on synced lips.

5. Image-to-video product turntable

[Start frame: a clean product photo of a running shoe on a white background]
Add motion: the shoe rotates slowly clockwise on an invisible turntable while the studio reflection shifts across its surface. Keep the shoe, colors, and logo exactly as in the source image. Medium close-up, static camera. Preserve the clean white background and soft shadow.
SFX: a subtle low hum.
Music: minimal upbeat electronic.
(1:1, 1080p, 6s)

Best for: turning one catalog photo into a rotating clip with accurate branding. Why it works: anchoring "keep the logo exactly as in the source" stops Wan from redrawing the product while it adds motion.

Want more finished examples before you start? Browse the 35 best Wan prompts, read how to prompt Wan for realistic video, or grab ready-made skeletons from Wan prompt templates.