Wan 2.2 Prompts

A Wan 2.2 prompt is a direction, not a description. The model already has a scene — either the one your words imply or the still image you supplied — so the useful work in a prompt is telling it what should happen and how the camera should behave, rather than recounting what is visible.

The three things a prompt should name

Most prompts that disappoint are missing one of these, and it is usually the third:

1. The subject — who or what the shot is about.
2. The action — what it is doing, in plain verbs.
3. The camera — how the shot moves. "Slow push in", "static wide", "handheld follow". This is the one people leave out, and it is the one that most changes how the result reads.

"Slow push in on a woman in a red coat walking through a rainy street, shallow depth of field" gives the model all three. "A woman in a red coat" gives it one, and a subject with no direction tends to drift.

Be literal, and use the negative prompt

Concrete nouns and named lighting beat stylistic adjectives. "Overcast afternoon light" changes the result more than "cinematic", because the model can act on the first and can only interpret the second. Name the lighting, the framing and the pace, and leave mood words for last.

What you do not want goes in the negative prompt, not the positive one. Suppressing blur, extra limbs, warped hands or burned-in text by name works far better than adding more description to the positive field — piling on positive clauses tends to dilute the instruction rather than sharpen it.

Iterating without wasting renders

Prompt iteration is where most of a budget quietly goes, so treat it like a small experiment:

• Change one thing at a time. Prompt and seed together tells you nothing about either.
• Fix the seed to compare wording. Same seed, two prompts, look at both. That isolates the wording.
• Iterate at a low resolution band, then re-run the winners at the size you will publish. See Wan 2.2 Resolution for why the band matters to cost.
• Keep the prompts that worked. A prompt that produced a good clip is a reusable asset, and re-deriving it later costs more than writing it down.

If you are animating a still image rather than generating from scratch, the prompt narrows further — you describe only the movement, since the composition is already fixed. Wan 2.2 Animate covers that case.

What a prompt cannot fix

No wording compensates for a source image that is too soft, a movement so large the model has to invent most of the frame, or a route that was not meant for the job. When a shot keeps failing, change the input or the route rather than adding clauses: text to video builds a scene from a description, image to video starts from a frame you supply, and extend a video continues a clip that already exists.

Cost depends on the model, resolution and duration you choose and is shown beside the form before you submit. Credit prices are on pricing.

Questions people ask

How long should a prompt be? One or two clear sentences naming subject, action and camera. Length is not the lever; specificity is.

Do style words help? Somewhat, but named lighting, framing and pace help far more.

Should I write prompts in English? English is the safest choice if you are unsure, since prompt interpretation is broadest there.

Where do I start? Open the Wan workspace and choose the Wan 2.2 route you want, then iterate at a low band before committing.