Blue gradient illustration with a central rounded outline

Wan 2.2 Prompt: Describe One Action and Keep the Settings Separate

Write a focused Wan 2.2 video brief with subject, action and camera direction. Verify the exact model and input mode before testing it.

Frequently Asked Questions

What complete prompt or practical brief can I copy?

Original video prompt: a small white sailboat drifts slowly across a calm harbor at early morning. A single adult in a blue jacket sits still near the stern. The camera makes a gentle side-tracking move at a constant distance. Soft warm light, stable sail shape and believable reflections. Keep one boat, one person and consistent shoreline geometry. No storm, sudden turn, text or scene cut. Use an available short duration and record the actual Wan 2.2 model, input mode and output settings separately from this descriptive prompt.

Where should I verify Wan 2.2 information?

Use the publisher’s Wan 2.2 repository and the current documentation of the service you actually use. A model repository and a hosted interface can expose different workflows and settings. Keep the exact model label and input mode with your test. This guide gives original descriptive prompts, not a promise of every provider’s duration, resolution or free allowance.

What is a good prompt structure?

Start with the subject, setting and one action, then describe framing and camera movement. Add details that must remain stable and remove conflicting events. In the harbor example, the boat drifts, the person stays seated and the camera tracks gently. Those roles are explicit, which makes the output easier to inspect than a request for a generally cinematic sea adventure.

What should I do if the boat or person changes?

Simplify movement and composition. Try the fixed-camera variation to see whether environmental motion alone is stable. Keep the subject count and clothing clear. Inspect the middle frames, where a model may reinterpret the sail or face during movement. If the action is complex, create shorter shots and assemble them rather than forcing the entire sequence into one unstable generation.

Should I add captions and music in the prompt?

If the selected model and service support audio, use their documented controls; otherwise do not assume prose alone enables it. Exact captions are often easier to add afterward as editable text. Keep raw generation separate from post-production when comparing results. Use authorized sound and inspect the actual export so a good visual preview does not hide a missing or mismatched audio track.

Write a Prompt You Can Evaluate

Keep the scene focused and verify the exact mode and settings.