AI video wind prompts: show what the breeze moves
Describe the visible response to wind: which hair, cloth or leaves move, where they lean, and whether the breeze stays gentle or becomes a gust. Give the camera a separate instruction. Start with one clear response before coordinating a whole landscape. These words express a visual target; they do not set a physical wind field or guarantee a stable face.
1. Make wind observable
A request for “a windy atmosphere” leaves the important movement open. A useful brief names a responder: loose hair lifts toward the right, the free end of a scarf flutters, or the outer leaves of a shrub sway. Choose movement that your composition can actually show. A close crop around a face cannot demonstrate what the unseen trees are doing.
Runway's Gen-4 prompting guide separates subject, scene and camera motion, and recommends starting simply before adding details. Google's video prompt guide includes weather and wind-blown leaves as scene context. The planning method below applies those ideas to a reviewable shot; its examples are our own, rather than documented controls shared by every model.
- Direction: for a fixed view, describe visible travel toward frame left or right. Avoid an ambiguous “left wind.”
- Intensity: say whether a few strands lift or a loose edge repeatedly flutters. “Strong wind” alone leaves the response unspecified.
- Progression: choose a steady breeze, one rising gust, or a movement that settles. Avoid asking for all three at once.
- Stable details: name the face, seated pose, bench or wall that should remain readable while flexible details move.
A scarf, short hair and a sheltered shrub need not swing by the same amount or in perfect synchrony. Look for a coherent visual impression in the intended scene. If the camera turns, screen direction can change too. Use a fixed view for an initial direction check, then plan a camera move as a separate decision.
2. Copy a wind-shot card
Visible responder and opening position: Direction in this camera view: Small breeze or larger gust, described as visible motion: Steady / rising / settling progression: Details and structures to preserve: Camera view and allowed movement: Reference file, selected model/mode and exact prompt: Output ID and observed motion / continuity: Accept or reject, with the visible reason:
Our original example is a seated adult with shoulder-length dark hair and a loose rust-colored scarf on a gray courtyard bench. A small shrub stands behind the bench, with a plain brick wall farther back. Keep the composition simple enough to see the scarf edge, face and bench line.
On a small screen, scroll the table sideways.
| Element | Intended response | Review question |
|---|---|---|
| Loose scarf end | A small flutter toward frame right | Does it stay attached and retain its color and width? |
| Hair | A few loose strands lift in the breeze | Does the face remain readable as the hair moves? |
| Shrub | Light supporting movement in the outer leaves | Do leaves remain part of the shrub rather than dissolve? |
| Bench and wall | Stable structure in the fixed view | Do straight edges bend, drift or pulse with the wind? |
3. Adapt these original wind prompts
These are unrendered drafts, with no benchmark or success-rate claim. Check the chosen service's actual input mode and duration options. Begin with a short trial that can answer one question before spending on a longer or more complex take.
Text-to-video: establish the courtyard scene
Locked medium view of an adult with shoulder-length dark hair and a loose rust-colored scarf, seated on a gray courtyard bench. A small shrub sits behind the bench in front of a plain brick wall. A gentle breeze lifts the free end of the scarf toward frame right. A few loose strands of hair and the shrub's outer leaves move lightly in the breeze. The person holds a relaxed seated pose. The bench, wall and soft daylight stay steady.
This version establishes appearance and surroundings as well as motion. If several responders make the result hard to judge, start with the scarf alone. Add the hair or leaves only after reviewing that attempt. A longer description does not enforce a shared simulation across them.
Image-to-video: use what the reference already shows
The free end of the scarf flutters gently toward frame right. The person holds a relaxed seated pose and the camera remains locked. The bench and wall stay steady.
Use this when the actual reference already contains the person, scarf, bench and wall. Attach it through the service's supported image input; mentioning a filename does not upload a reference. Runway's Gen-4 guide treats the input image as visual context and recommends focusing the text on motion. Do not ask a tightly tucked scarf to move like a long free end without changing your reference or intention.
Fabric-only alternative: one gust and a settling edge
Fixed close view of a pale linen curtain beside an open doorway. One brief gust lifts the free lower edge inward, then the cloth settles into a gentle sway. The curtain remains attached along its top edge. The doorway and camera stay steady.
The opening, lift and settling are intended beats, not frame-level timing commands. This alternative also avoids needing a face in the shot. It does not prove the model has solved cloth physics or make a cropped-out problem correct.
4. Review the movement and what should stay still
- Watch the whole take. Check whether the intended breeze reads at normal delivery speed before pausing on details.
- Follow the responder. Look for a scarf detaching, fabric changing shape abruptly, hair merging into a cheek, or leaves disappearing.
- Compare flexible and rigid parts. In a locked view, inspect the wall and bench for warping while the scarf moves.
- Check direction and progression. Note unexplained reversals or a sudden storm when the brief asks for a gentle breeze.
- Inspect the ending. Keep a take only if its final motion and visible details suit the intended edit.
Save a concrete observation such as “the scarf narrows and detaches during the lift” or “the wall bends while the camera is fixed.” Motion blur can conceal a defect without resolving it. Record the input, prompt, settings and output ID; a single clip does not reveal the model's internal cause. Use the character consistency guide when hair movement also changes the person's appearance.
5. Change the next attempt for a visible reason
If the motion is absent, check that the reference exposes a movable edge, then make that response more explicit. If everything deforms, reduce the requested motion or remove supporting responders. If the camera makes direction difficult to judge, try a fixed view. Keep the other recorded inputs unchanged so the revision is interpretable; stochastic generation still means results can vary.
Set an attempt limit. When exact garment behavior or a particular performance remains essential, consider suitable source footage or documented animation and editing controls. For an atmospheric insert, a simpler B-roll brief may serve the edit. A repeating breeze needs a separate loop and seam check; wind movement alone does not make a seamless loop.
Shotmelo currently helps organize manual prompts, preview a local reference, and save or export text drafts. It does not render video, simulate wind or cloth, analyze motion, or repair footage. Use the card to prepare your brief, then review the actual take in your chosen generation or editing service.
Published November 3, 2026. Official guidance checked on that date. Card, scene, prompts and review criteria are original planning examples; no paid rendering tests or guaranteed physical result.