AI scene animation

#Animation

The Animation block sits in Layout mode, under the elements you arranged. From top to bottom it holds the Motion In / Motion Out toggle, a prompt box, a row of variable chips, the preset chips, Generate, and once a motion exists, Clear and a read-only view of the motion that was written. A badge reading ANIMATED · IN or ANIMATED · OUT tells you at a glance that the scene carries a motion and which way it runs.

How it works

You describe the motion, and Wildfield writes it as a timeline for that scene. The timeline only moves, fades, scales, rotates and blurs elements. It never changes their layout, so the positions you set in Layout are always where an entrance lands and where an exit leaves from. It can split text into lines, words or single characters, which is what makes letter staggers and typewriter reveals possible.

The motion is written to be deterministic: no randomness at play time, nothing tied to the clock or the pointer. That is why the preview, the HTML export and the video render all show the same frames.

Only the elements you name in the prompt move. Anything you leave out stays exactly where it is and fully visible, so "pop the CTA in" leaves the headline alone.

Variables

Name elements with variables in curly brackets. The chips under the prompt box insert them at the cursor, and a chip only appears when that element exists in the scene:

• {bg} the background (always offered)
• {h1} the headline, {h2} the subhead and {h3} the tertiary text
• {cta} the CTA button
• {logo} the logo, when the design system has one
• {disclaimer} the disclaimer
• {subtitles} the subtitles, when the scene has subtitle text
• {duration} the animation window

Timing in the prompt runs from 0 to 1 across {duration}, where 1 is the whole window. The window is the scene length minus any Hold, and the chip shows it in seconds (for example {duration} · 1.0 (4s)), updating live when you change either. So "pop {cta} in at 0.2" means a fifth of the way through, whatever the scene length. Vector layers are not offered as variables: they move with the canvas tools instead.

Presets

The preset palette is the fastest way in. Clicking a chip replaces the whole prompt with a ready-written description, and it never generates on its own, so you can browse and read presets for free. The Generate button glows to remind you to press it. Presets are written about "each element" or "the headline" rather than a specific scene, so any preset works on any scene.

There are 36 presets each way. On Motion In they are grouped as entrances (Fade in, Slide up, Pop / scale, Blur in, Flip in, Rise & settle and more), text reveals (Typewriter, Word by word, Letter stagger, Line by line, Wave, Bounce letters, Elastic letters), reveals and masks (Wipe reveal, Mask up, Clip reveal, Curtain), rolls and kinetic type (Slot roll up, Odometer, Snappy roll, Glitch, Shimmer, Kinetic pop) and specials (Count up, Highlight sweep, Hero build, Cascade). Motion Out mirrors them as exits: Fade out, Zoom out, Drop & fall, Shrink out, Untype, Wipe out, Mask down, Curtain close, Glitch out, Dissolve, Count down, Hero teardown and the rest.

A preset is a starting point. Edit its wording before you generate: swap "the headline" for {h2}, change "staggered" to "all at once", or add a second beat.

Generate, iterate, clear

1. Pick Motion In or Motion Out.
2. Write a prompt or click a preset, then edit it.
3. Press Generate. The button reads "Generating..." while a bar sweeps under it. Generating uses credits.
4. The motion is applied straight away, with a toast saying Motion In applied or Motion Out applied. Press Play to watch it.
5. Not quite right? Edit the prompt and generate again. The new motion replaces the old one, and the prompt stays in the box so you only change what you need.

Clear removes the motion but keeps your prompt text, ready for another try. Generate, Clear, a preset click and a direction change can all be undone. The written motion is shown read-only: you shape it through the prompt, not by editing it directly.

One prompt, every format

The motion belongs to the scene, not to a format. The same prompt drives the 9:16, the 1:1 and the 16:9 versions, because it names elements rather than coordinates, and each format's elements already sit where you placed them for that ratio. Duplicating a scene copies its animation too.

Tips

• Be specific about order and timing: "Fade {h1} in over the first 0.3, then stagger {h2} letters from 0.3 to 0.6, pop {cta} in at 0.7" gives far more control than "animate everything".
• Say how elements should arrive: "from below", "from the left", "with a slight overshoot", "blurred to sharp".
• The paused editor and PNG exports show the rest pose, without the motion, so a still never catches an element half faded.

#Motion In

Motion In is an entrance. Each element you name starts hidden or offset and ends at rest, exactly where you placed it in Layout. An entrance never leaves an element invisible at the end, so a finished Motion In always reveals the full layout.

When it plays

Motion In is anchored to the start of the scene and fitted to the animation window. If the scene has a Hold, the entrance plays first and the Hold adds still frames after it, so the finished frame sits on screen before the cut. How long each element takes inside the window is up to your prompt: "over the first half" and "quickly, in the first 0.2" give very different pacing.

Because the motion is written in window units (0 to 1) rather than seconds, changing the scene length later re-fits it proportionally, with no new generation and no extra credits. A 3 second entrance stretched to a 6 second scene plays at half the speed, beat for beat.

Stagger and per element motion

Staggers come from the words you use. "Stagger the letters", "word by word" and "one after another" all produce offset timing, and most of the text presets are staggered already. Different elements can do different things in one prompt: the headline can wipe in while the CTA pops.

Worked example

A 9:16 story ad for a coffee subscription, 5 seconds long, with a headline, a subhead and a CTA.

1. In Layout, place the headline in the top third, the subhead under it and the CTA near the bottom, inside the safe area.
2. Open Animation, keep Motion In selected and click Word by word.
3. Edit the prompt to: "Reveal {h1} word by word from below over 0 to 0.4, fade {h2} up from 0.35 to 0.6, then pop {cta} in at 0.65 with a small overshoot."
4. Press Generate and play the scene. Everything lands by 0.75 of the window and the ad reads cleanly for the rest.
5. Switch to 1:1. The same motion plays on the square layout without another prompt.

#Motion Out

Motion Out is an exit. Each element you name starts visible at its Layout position and ends hidden, off screen or scaled away. Use it on scenes that hand over to something different: the last scene before a hard cut, or a scene that clears the stage for a product shot.

When it plays

Motion Out is fitted to the same window as Motion In, but it is pushed back by the Hold, so the ad sits still first and the exit runs at the end of the scene. How close the exit lands to the cut depends on your prompt: ask for the exit "in the last 0.3" or "from 0.7 to 1" so it finishes on the cut rather than partway through the scene.

One direction per scene

A scene carries one motion: Motion In or Motion Out, not both. The toggle picks which kind the next Generate writes, and flipping it keeps your prompt text and swaps the preset palette. If you want an element to enter and then leave, split the idea across two scenes: an entrance on one, an exit on the next.

Linked scenes

Scene motion uses flat 2D moves only, so a settled entrance matches the next linked scene pixel for pixel. Each scene in a linked family runs its own motion inside its own window, which lets you chain an entrance on the first scene into an exit on the last.

Worked example

A three scene 16:9 ad for a gym opening.

1. Scene 1 introduces the offer with a Motion In (Slide up on {h1}, then Fade in on {cta}).
2. Scene 2 shows the location and has no text motion.
3. Scene 3 is the end card. Set a 1 second Hold, choose Motion Out and click Curtain close.
4. Edit the prompt so the curtain closes over {h1} and {cta} "from 0.7 to 1" and leaves {logo} in place.
5. Generate and play from scene 1. The copy sits still through the Hold, then the curtain clears it and only the logo remains as the ad ends.