Wan AI Video Prompt Guide: Turn a Simple Idea Into a Usable Shot

How should you write a Wan AI video prompt?

Write the prompt as a shot brief. Start with what the viewer should see, describe what changes during the shot, then tell the camera how to observe it. This makes the result easier to inspect and easier to revise.

The method in this guide is designed for the publicly documented Wan family, including workflows built around Wan2.2 and Wan2.7. Wan 3.0 does not yet have a verified official public specification, so these examples should be treated as portable prompting patterns rather than confirmed Wan3.0 features.

The seven parts of a strong video prompt

You do not need a long paragraph. You need the right information in the right order.

The subject and action should appear early. If the most important instruction is buried after a long style description, it becomes harder for you to tell which part of the prompt caused a weak result.

A text-to-video prompt template

Use this structure when you are creating the scene from words.

[Shot type]. [Subject] [main action] in [setting].
[Camera movement]. [Lighting and visual treatment].
Keep [important detail] stable. [Sound or dialogue, if needed].

Here is a product example.

Close-up product shot. A red ceramic cup sits on a dark wooden table in a quiet kitchen at sunrise. Steam rises gently from the coffee while the cup turns five degrees toward camera. Slow dolly in, soft window light, shallow depth of field. Keep the cup shape and white logo unchanged. Ambient room tone, no music.

The prompt gives the shot one job. The cup moves slightly, the camera moves slowly, and the surrounding scene stays quiet. That gives you a result you can judge.

Here is a narrative example.

Medium shot. A tired courier stops beneath a neon sign on a rainy street and looks at the unopened envelope in his hand. The camera tracks gently from left to right. Wet pavement reflects red and blue light, restrained cinematic color, realistic rain. Keep the courier's dark jacket and canvas bag consistent. Distant traffic and footsteps, no dialogue.

The exact wording will vary by model and provider. The useful part is the separation between what happens, how it is filmed, and what must remain stable.

A practical image-to-video prompt template

An input image already carries much of the visual identity. The prompt should spend more space on movement and less space trying to redraw the subject.

Animate the provided image. [Subject] performs [one main action].
The camera [one camera movement]. Preserve [shape, face, clothing, or product detail].
Use [lighting and motion quality]. [Sound or dialogue, if needed].

For a product image, try this.

Animate the provided product image. The bottle rotates slowly from a three-quarter view toward the front while a small highlight moves across the glass. The camera remains mostly static with a very slight push in. Preserve the label layout, bottle proportions, and cap color. Clean studio light, soft background, controlled motion. No text changes and no extra objects.

For a character reference image, try this.

Animate the provided character image. The character takes two slow steps forward and looks toward the light on the right side of frame. The camera tracks backward at walking speed. Preserve the hairstyle, coat, face shape, and color palette. Overcast afternoon light, gentle wind in the coat. No new characters and no sudden camera shake.

The image-to-video approach is useful when the starting appearance matters. The image handles the visual baseline. The prompt handles the next movement.

How to fix common Wan prompt failures

The output contains too many actions

Reduce the shot to one main action. A person can open a door and look inside. Asking the same person to run, sit, wave, turn around, speak, and pick up a phone gives the model several competing priorities.

The camera does something different

Write the camera as its own sentence. Use one primary movement such as a slow push in, a lateral track, or a gentle pan. A static camera is also a valid choice when the subject carries the scene.

The subject changes shape

Repeat the details that matter. For a product, name its silhouette, label position, and color. For a character, keep the clothing, hairstyle, age range, and defining accessory consistent. Do not add a new description every time you regenerate.

The scene feels empty

Add a few physical details that affect the shot. Mention the surface under the object, the direction of light, the background depth, or the sound source. Three useful details are better than a page of mood words.

The result looks polished but misses the idea

Move the core action to the first sentence. Style can support the shot, but it cannot replace the shot's purpose.

A prompt revision loop that saves time

Try a Wan video prompt

Use the Wan video generator to test a simple shot first. Start with one subject and one action, then bring the prompt back here when you need to revise a specific failure.

Related Reading

• Wan 3.0 Official Release Status — Separate confirmed Wan information from third-party model names
• Wan AI Video Workflow — Turn several prompts into a coherent short video
• Wan2.7 video generator — Explore a current Wan video workflow

FAQ

Disclosure

The prompt structures and examples in this article are editorial templates for Wan-family video workflows. They are not presented as official Wan3.0 specifications or as a record of a controlled model benchmark. The model-status references were checked against the Wan-Video GitHub organization, the Wan-AI Hugging Face organization, and Alibaba Cloud's Wan2.7 API reference on August 6, 2026.