Every job starts with a reference and the same gap: you can see it, now build it. Prism closes the first hour of that — the layout, the layer stack, the timing — so you start from something moving instead of an empty comp.

How it goes

Describe what you are looking at

Structure, not adjectives. "Full-bleed dark background, a headline that rises in from below, a card behind it that pops, logo bottom right, everything settles by two seconds."

Let it build the whole scene

Prism creates the comp, the layers, the layout and the timeline in one call, then returns frames of the result so you can see what it made without switching windows.

Correct it in words

"The card is too fast, and the headline should land before it." It edits the same scene rather than starting over.

Take it the rest of the way by hand

What you have is native layers and keyframes. The reference got you to a working draft; the taste is still yours.

This is the part that has no free equivalent

Importers and scripts move assets around. Building a scene from a description — measuring text, laying it out, sequencing it — needs something that can reason about the whole comp at once.

What it is good at, and what it is not

Good at	Still yours
Layer structure and stacking order	The idea
Layout that respects real text bounds	Typography choices worth arguing about
Sequencing and stagger	Timing that feels right rather than correct
Rebuilding a variant fast	Knowing which variant to keep

Try it

"Build a title card: dark gradient background, headline rising from below, subtitle fading in 0.3s after, both centred."
"Same scene but the headline pops instead of rising, and everything is 20% faster."
"Show me frames at 0.5s and 2s."