Collage
The speaker as a paper cutout, with text slamming in on each word
Turn a clip of a teacher, coach or expert talking to camera, or a written script with no footage, into a 35-90 second vertical ad or short lesson (1080x1920) in a paper-collage look.
Important: this is a skill for terminal AI coding agents (Claude Code, Codex) that runs on your own computer. It is not a prompt for the ChatGPT website. Check the requirements below before buying.
What it makes
The speaker is cut out as a white-edged sticker on a cream paper plate with a torn-paper sun and clouds. Key words slam in on the spoken word as big yellow and orange type with a hard shadow. Pains and ideas are illustrated with vintage statue and sepia newspaper cut-outs, lists become yellow highlight bars, numbers become big orange numerals.
Two sources:
- Footage (default): your clip of one person speaking is cut, framed 9:16 and the background removed.
- Faceless: no footage. The agent writes and voices the script (VieNeu, free, or ElevenLabs) and the scenes use a statue "host" and cut-outs only.
The story of the video
Pick one of two shapes:
- Ad (course, service): reverse hook, pains, the offer named, how it works, proof, urgency, sign-off.
- Lesson or insight: a claim or question, the common belief, the turn, an example with numbers, the punchline, a plain-subtitle close.
Cut-out scenes take about 25-35% of the runtime; the rest is the presenter.
How it works
Open your agent next to your clip and ask, for example:
Làm video collage giấy từ clip giáo viên này, dạng quảng cáo khóa học, khoảng 60 giâyThe agent transcribes the whole clip once, picks 35-90 seconds of one idea, cuts out fillers and dead air, removes the background, writes the edit and checks that every keyword lands on its spoken word. Captions never invent facts: every caption word is something the speaker says, and numbers on screen are numbers spoken.
What you get
- The skill folder with its template, scripts and a bundled art library: the collage plate, 3 backgrounds and 15 vintage cut-outs.
- A script to generate extra cut-outs in the same style when you need a new picture.
What you need
- Claude Code or Codex, with Python and
npx hyperframesfor checks and rendering. - A speech-to-text key:
SONIOX_API_KEY(default) orELEVENLABS_API_KEY. - Optional: a Key4u or
GEMINI_API_KEYkey for extra cut-out art. - Time for the background removal: about 0.7 seconds per frame on CPU, so roughly 20 minutes for 60 seconds.
Install and make your first video
Copy the skill folder into your skills folder:
~/.claude/skills/talkinghead-collage-video/For Codex, use ~/.agents/skills/talkinghead-collage-video/. Then open the agent in a project folder and point it at your clip.
Who it is for
A teacher, coach or expert who wants a talking-head clip, or a voiced script, turned into a paper-collage ad or short lesson.
Not a fit if you want
- A line-diagram explainer.
- A plain cut with no graphics.
- A long recording kept whole.
Frequently asked questions
Does it generate video with AI?
No. Only still cut-outs, the voice in faceless mode, and the bundled music and sound effects. Generated people are fictional only.
Can I make a video without filming myself?
Yes. Faceless mode writes and voices the script, then uses the statue host and cut-outs instead of a person.
Can I use the videos commercially?
Videos you render are yours for any commercial use. You may not resell, redistribute or share the skill files.
More looks from Talking Head
Full set: Talking Head
3 looks

Diagram
A line diagram growing beside the speaker's face bubble

Edited short
A carefully edited short: grade, captions or keywords, title behind the head
Included in ReelCrewHighlight
A highlight short with motion-fx design




