Diagram
A line diagram growing beside the speaker's face bubble
Watch on YouTubeTurn a clip of someone explaining one idea, or a written script with no footage, into a 45-75 second vertical explainer (1080x1920) where one line diagram grows on the spoken words.
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 look is minimal line diagram on pure black: soft rounded outline boxes with a faint glow, curved arrows that draw on, green numbered step rings, a cycle that closes, a plus and minus ledger with line icons. The speaker sits in a small 300 px window (rounded square by default, or a circle) and cuts to a full-width band on one or two key lines.
One video is one diagram that grows. There are no scene changes: finished parts dim, or the camera reframes. The agent picks one base and extends it:
- Flow: a box, arrows, more boxes.
- Cycle: numbered steps that close the loop.
- Ledger: plus and minus columns with icons and totals.
- Growth or compare: a counter on an arrow, or two boxes side by side.
A takeaway pill near the end, then a brand outro. Talks of 3-8 minutes are split into chapters, one short diagram each.
How it works
Open your agent next to your clip and ask, for example:
Làm video sơ đồ nét từ clip này, người nói ở ô tròn góc trênThe agent transcribes the clip, picks 35-75 seconds of one idea, frames the speaker, designs the diagram from the transcript (6-14 items), cues each item on the word that introduces it, then checks and renders. The diagram only shows what the speaker says: every box, label, number and step is a word or number spoken in the same sentence.
What you get
- The skill folder with its template, scripts and line icons.
- A timing table and layout warnings before every render, so nothing lands under the speaker window or the captions.
- A faceless mode: the agent writes and voices the script, and captions move to the bottom.
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. - Faceless mode: VieNeu (free) for Vietnamese; other languages use ElevenLabs only.
Install and make your first video
Copy the skill folder into your skills folder:
~/.claude/skills/talkinghead-diagram-video/For Codex, use ~/.agents/skills/talkinghead-diagram-video/. Then open the agent in a project folder and point it at your clip.
Who it is for
One idea explained by a speaker as one line diagram that grows on the spoken words, with the speaker in a small bubble.
Not a fit if you want
- A collage ad.
- An edit heavy with b-roll.
- A plain cut, or an explainer with many scenes.
Frequently asked questions
Does it use AI images or AI video?
No. Everything on screen is HTML, SVG and bundled line icons. The only pictures allowed are your own screenshots or photos, shown while you talk about them.
Can I use a long talk?
Yes. A talk of 3-8 minutes is split into chapters, each one short diagram of 6-14 items about one idea.
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

Collage
The speaker as a paper cutout, with text slamming in on each word

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




