See the comp: frames back to the agent

Snapshot is how the agent sees. It renders a frame of your composition and gets the image back in the same reply: no files, no paths, nothing to grant. That is what turns "I made the change" into "I made the change and it looks right". The agent uses it after any visual change, together with a script so it can change and see in one ask, and zoomed in when a label is too small to read.

How to use it

You say: "Show me the title card at two seconds." · "Fade the date in over half a second and show me." · "Zoom in on the chip bottom left so you can read it."

You rarely need to ask. A good agent snapshots on its own after it builds or changes something, and tells you what it saw.

What you get

Back to the agent: the image, its size and the time it was taken. With a script, the frames sit beside the script's result, so a mistake is visible before the agent reports back. A scene returns its own frames the same way.

In the project: nothing changes. A snapshot is a render, not an edit.

Notes and limits

• Time. A frame takes seconds, a long comp more. Ask for one or two frames per request; the agent knows to.
• Resolution. Frames render at a review size, never at the viewer's half or third setting, so what the agent sees is what you would see.
• Stale frames. After Effects can serve a cached frame from before the last edit. Prism clears the cache when a change preceded the snapshot; if a frame still looks old, ask for it again.
• Busy After Effects. An open dialog holds the render until it times out; see After Effects busy.
• Active comp. Rendering another composition can leave it in front afterwards; bring yours back before the next request. This and the occasional stale frame are on known issues.
• Cost. A snapshot is an MCP action; checking on a slow call is not.

Common questions

Related