← All skills
Talking-head Video
Your take cut clean, the edit drawn on your own words, then into 剪映 or CapCut.
Talking-head Video: its example films and styles →
How it works
- 1Copy the install prompt and paste it into your agent.
- 2Your agent runs mo doctor to check Node.js, FFmpeg and Chrome, and tells you how to install anything missing.
- 3Give it your material. It renders an MP4 on your computer, and you ask for changes in plain words.
Paste into Claude Code, Codex, Cursor or any agent that supports skills.
Films it made
All it takes
No API keyThis is all it needs
- Video or a recording
Dependencies
Node.js 22+, FFmpeg, Chrome, and whisper-cpp with its speech recognition models (about 600 MB). No API keys.
Before it starts, your agent runs mo doctor to check Node.js, FFmpeg and Chrome and tells you how to install anything missing; if this skill needs more (such as whisper for captions), its instructions say so.
Example prompt
Use the talking-head-director skill to cut take-03.mp4 clean, caption it, and send it to 剪映.
Style prompt (Claude Motion or Opus 5.5)
Edit my talking-head recording [file] into a [length] video: cut it clean (show me the cut plan first), proofread the captions against the recording, put my name and the question in the first seconds, cards on my own words that never cover my face, then export a 剪映 draft for [who will watch, e.g. people who follow our channel]. Use only my own recording and the words I say in it; don't change a word or number. End on the last line I say, with the caption held for 2 seconds.
Nothing to install: paste this into Claude Motion (Team or Enterprise) or give it to Opus 5.5 for a similar look. The result changes from run to run; the skill works from your own material and renders on your computer.