Skip to content
Prism
Use cases

Talking head video in After Effects with AI

You can edit a talking head video in After Effects with AI: ChatGPT, Claude or any MCP client connected to Prism transcribes the take, then builds captions, icon cards, cuts and punch-ins in your open comp as editable layers.

The YouTube edit is a recipe: cut the pauses, punch in on a new sentence, caption every word, land a graphic on the keyword, add a whoosh. Every tool that does it renders an MP4 you cannot open; here every piece stays a layer.

What lands

Built in After Effects through Prism from real footage, as a sequence of asks, not one prompt:

ElementWhat it is in the comp
Captionsone text layer per word group, timed to the transcript, at chest level, the keyword in a second colour
Icon cardsnear-black cards with a white icon, rising to land on the spoken word with motion blur and a shimmer sweep, then floating out
Cuts and reframesthe footage split at the pauses, punch-ins as scale and position keyframes, a snap on each new sentence
Soundsound effects on cue markers, one per card and cut

What you say, and what lands

You typeWhat appears in your comp
"Captions at chest level, three words at a time, Inter-Bold, white with a dark stroke, the keyword in orange."One text layer per group, in and out on the words, a text animator on the keyword
"On 'audience', 'budget' and 'seven', land a black card with a white icon on the word: rise in over 9 frames with an 8% overshoot, motion blur on, a shimmer sweep, hold, then drop."Three shape-layer cards with icon precomps, position keyframes, comp motion blur, a Merge Paths shimmer
"Cut every pause longer than 0.4 s and punch in 12% at the start of each sentence."The footage split on the gaps between words, scale and position keyframes on each punch-in
"A whoosh on every card and a click on every cut."Audio layers placed at the cue markers

Test it on your own project. A Free account is enough to start: Sign up, then connect your AI and ask for this.

How to do it

1
Transcribe the take
Select the footage layer and ask for a transcript. A [Transcript] guide layer with a marker per word appears above it.
2
Ask for the captions
Words per group, position, font, stroke, and what the spoken word does. That sentence is the style.
3
Land icon cards on the keywords
Name the words and the look. Each card starts a few frames early so it lands on the word.
4
Cut and reframe
Say the pause length and the punch-in size. Cuts fall on word boundaries; punch-ins are keyframes on the footage layer.
5
Add sound
Import the effects you have and ask for them on the cues, or generate one at the playhead, paid per generation.

YouTube, Shorts and LinkedIn

StyleFrameCaptionsGraphicsCutting
YouTube16:9chest level, three or four wordscards in the empty thirdpauses cut, a punch-in per sentence
Shorts9:16, face lower so the top third is clearbigger, one to three wordscards above the head, never over the facetighter, every gap over 0.2 s
LinkedIn1:1 or 4:5full sentences, calmera title card and a logofewer cuts, no snaps

What it needs

After Effects2024, 2025 or 2026, macOS or Windows
AI clientany MCP client: ChatGPT, Claude, Claude Code, Cursor, Codex and others
PlanPro with a Prism AI balance for the transcription; captions, cards and cuts use MCP actions

What it does not do

It does not remove silences on its own: a pause is cut where you ask, using the gaps between words in the transcript. Nothing tracks the face; a punch-in is scale and position keyframes around a point you name.

Common questions