agentleFS
WEBGRSWEBGRS/openfilm-talking-head/skills/talking-head/SKILL.md

talking-head

Skill0 starsChanged 14 days agoMITView on GitHub

Cut a recording of someone talking to camera into the version they would give on a good day - ums, false starts, repeats, retakes and dead air gone with cuts nobody notices, their key words on screen as they say them, a name card and a cover. Use when the person gives a talking-head recording, a vlog, a course or a long rough take and asks to clean it up, tighten it or cut it down, or to add key words, a name card or a cover (口播剪辑, 剪口播, 去口癖, 去气口, 粗剪, 精剪, 重点词, 人物名片, 封面).

6 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions6 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 6

    1. sanky369/vibe-building-skills★ 30same name
    2. ofoxai/skills★ 2same name
    3. OpenArt-AI/skills★ 1same name
    4. vidmoat/skills★ 1same name
    5. puntorigen/skills★ 0same name
    6. zuckenbert/video-from-code★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No one can tell where it was cut, and what a viewer would skip is gone.

    line 13
  2. Distinctive:

    Their words, in their order: Only ums, false starts, repeats, abandoned takes and dead air go, unless they ask for more.

    line 18

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Talking head

Cut the person's recording into the version they would give on a good day: the same person and their own words, with nothing a viewer would skip and nothing missing that mattered.

What defines it

No one can tell where it was cut, and what a viewer would skip is gone. A cut that clips a syllable or swallows a breath is heard at once, and a sentence spliced from words they never said in that order is no longer theirs. Taking out every pause is not tightening: breath and emphasis stay.

What has to be true

  • Their words, in their order. Only ums, false starts, repeats, abandoned takes and dead air go, unless they ask for more; of a line said twice, the last whole one stays. A silence over 0.8 s is cut to about half a second.
  • Every cut falls in a pause of 0.12 s or more in the sound, never inside a word. A transcript's times run a few tenths off and it leaves ums out, so cut points come from the pauses measured in the recording. Where the picture would jump, something covers it; look at the frames on both sides of every cut.
  • What can come out does. They asked for the ums, restarts and dead air out. One that stands between two pauses is cut there, and check fails while it is in. One that runs straight into a word has no pause to cut in: it stays, and you tell the person how many and which.
  • Heard again, it is their words. The joined sound, transcribed once more, reads as what they said in their order, less only what was meant to go; a cut that left a repeat in or took a word is moved.
  • What is on screen is what they said. A key word lands as they say it, within 0.25 s, and leaves when they move on. The name card carries the name and title they gave, spelled their way, and goes in with what is sure (a name only a transcript heard is a guess: that part stays off until they spell it). Subtitles come from the transcript.
  • The cover is a line they said that says what the film is about, readable at the size a feed shows it, shown before the film plays, not a shot in it.

What you have

Their recording and its transcript with every word's time (npx openfilm get asr --src <file> writes <name>.vtt beside it). Keep the two together, in assets/ too: subtitles follow every trim from that file. node <this skill's folder>/scripts/cuts.mjs measures the sound, with the ffmpeg on the person's PATH:

  • pauses <recording>: every pause, with the words around it;
  • slips <recording>: every um, repeat, restart and long silence, with the cut that takes only it, or why it stays;
  • joined --out <file>: the clips back to back, to transcribe again; keep that file out of assets/ and delete it after;
  • check: run it last. Cut points, what is still in, order, the transcript, key words against the words, the numbers for your report. It exits 1 and ends with what to do.

Each key word, the card and the cover is a page of its own (a page's address takes no query); npx openfilm look <page> 0 gives the cover's picture. Transcribing costs the length of the recording in their service, and hearing the cut again costs its length once more: say so before you start.

With the person

What they ask for comes first. How long it should end up, where it will be watched, their name and title for the card, and whether tangents may go as well: ask without holding up the work. When done, tell them in a few lines the length before and after, the cuts over two seconds with their words so they can put any back, what stayed and why, any word taken that was not an um or a repeat, and what of theirs would make it better.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.