Talking Head: all looks
Talking Head
The three Talking Head looks sold on agentvid.io in one set: turn a clip of one person talking to camera, or a script for a faceless version, into a designed vertical short.
Important: these are skills for terminal AI coding agents (Claude Code, Codex) that run on your own computer. They are not prompts for the ChatGPT website. Check the requirements of each look before buying.
What is in the set
Each look is its own skill folder with its own guide, so you can use any of them on its own. All three make 1080x1920 vertical video.
Collage
The speaker as a white-edged paper sticker on a cream collage plate, key words slamming in on the spoken word, vintage cut-outs for pains and ideas. Story shapes for an ad or a short lesson, 35-90 seconds. Works from your clip, or faceless from a voiced script.
Diagram
One minimal line diagram that grows on the spoken words while the speaker sits in a small window, 45-75 seconds. Longer talks of 3-8 minutes become chapters. Nothing on screen is AI-generated; the diagram only shows what the speaker says.
Edited short
One real talking-head clip of 1-5 minutes polished into 40-90 seconds: word-level cut, natural grade, face-aware keywords and captions, a round speaker bubble over maps or demos, an optional Before/After hook.
Also in ReelCrew
Highlight, a highlight short with motion-fx design, is part of the same skill but runs inside ReelCrew; it is not sold on agentvid.io.
The set includes every edition of every look: Basic and Pro where a look has both.
Which one to use
- You sell a course or a service and want a bold ad: Collage.
- You explain one idea step by step: Diagram.
- You have a raw phone clip that just needs a clean, polished edit: Edited short.
About the caption looks
The caption looks (Karaoke, SRT and Phrase) are separate listings and are sold as they are. They run in the Fast cut pipeline on one recording of one person talking, so they cut silences and fillers as well as caption (karaoke also reframes to 9:16). They are not caption-only add-ons for a video you have already finished. They are not part of this set.
What you need
All looks need Claude Code or Codex and a speech-to-text key (SONIOX_API_KEY; Collage and Diagram also accept ELEVENLABS_API_KEY). Rendering runs through HyperFrames.
- Collage and Diagram: Python and
npx hyperframes. Faceless mode voices with VieNeu (free) or ElevenLabs. - Edited short: ffmpeg, Node.js 18 or newer, and a Python environment with
opencv-python-headlessandnumpy.
Install
Each look installs as its own folder in ~/.claude/skills/ (for Codex, ~/.agents/skills/):
~/.claude/skills/talkinghead-collage-video/
~/.claude/skills/talkinghead-diagram-video/
~/.claude/skills/talkinghead-edit-video/Who it is for
A designed short from one person talking to camera.
Not a fit if you want
- A fast plain cut.
- A long lecture.
Frequently asked questions
Do the captions add things the speaker did not say?
In all three looks the skill takes the numbers and claims on screen from what the speaker says (in Edited short, also from facts you supply with a source). Still read the video before you post.
Can I buy one look on its own?
Yes. Each look also has its own listing.
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

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

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




