Your first prompt
Once the panel is signed in and Prism's address is added to your AI client, you ask in plain language and the AI works in your live composition. This page shows a handful of starter prompts, what each one produces, and the one habit that makes results land on the first try. Start with the read-back below to confirm the connection, then move on to real edits.
The habit: select first, ask second
Prism reads the true state of your project (the active comp, its layers, their properties and keyframes) not a description you type. So the cleanest way to aim a prompt is to select the thing you mean before you ask. (This read-then-write loop is covered in how Prism works.)
Keep each request small and verifiable. "Animate this in" then "make the ease softer" beats one giant paragraph. You can check each result and steer the next step.
First, confirm the connection
Prompt: "Tell me about my current composition."
| You say | Prism does |
|---|---|
| "Tell me about my current composition." | Reads your active comp and describes it: name, dimensions, frame rate, duration, and the layers in it. |
What you get back: a plain-language summary of the comp that is open in front of you. If the AI reads it back correctly, you are connected and pointed at the right project. If it cannot see a comp, open one and make sure the panel is signed in and connected.
Animate a property
Prompt: "Scale the selected layer from 0 to 100% over the first 20 frames, with a soft ease."
| You say | Prism writes |
|---|---|
| "Scale the selected layer from 0 to 100% over the first 20 frames, with a soft ease." | Two keyframes on the Scale property (0% at the start, 100% at frame 20) with ease-out easing applied. |
What you get back: two real keyframes on the Scale property: one at the start, one at frame 20, with easing applied. Open the graph editor, drag handles, retime them: the output is editable AE keyframes, not a baked script.
Because the AI knows web motion vocabulary, you can also describe feel: "ease it like a gentle spring," and Prism turns that into keyframes.
Create a composition
Prompt: "Create a 1920×1080 comp at 30 fps, 10 seconds, called Intro."
What comes back: a new comp with exactly those settings, ready to build into. Follow up in the same breath ("add a centered text layer that says Prism"), and it scaffolds the layer for you to refine.
Import an SVG with real layers
Prompt: "Import this logo SVG and keep the text and shapes editable."
Prism's SVG import preserves real text and shape layers instead of flattening them to outlines, so you can recolor, retype, and animate the pieces after import. Pair it with select-first: bring it in, then ask to animate one shape at a time. If imported text looks thin or clipped, see troubleshooting.
AI-native: subtitles and beat markers
Two things AE can't do on its own:
| Prompt | What comes back |
|---|---|
| "Transcribe this audio layer into timestamped subtitles." | Timestamped subtitle text, placed on the timeline. |
| "Detect the beat of this track and drop markers on it." | Timeline markers on each detected beat, ready to snap layers to. |
These two are the only prompts here that spend your AI balance; everyday MCP actions don't. See usage and limits for what each capability costs and your monthly allowance.
When something looks off
If a result comes back wrong (thin or clipped text, an operation that didn't take, a font that won't apply) it's usually a known After Effects quirk with a one-line fix. See troubleshooting.
Related
- How Prism works: the read, reason, write loop and reversibility
- Real keyframes, not scripts: why the output stays editable
- Usage and limits: metering, monthly allowances, and where to see the current rates
- Troubleshooting: fixes for AE quirks and connection issues