HyperFrames by HeyGen
Official HyperFrames by HeyGen: the best AI video editor in chat. Captions, motion graphics, MP4.
Cinematic launch films and motion design from code, directed by your coding agent.
제작자라면 이 코드를 README에 붙여 넣으세요. 배지는 이 페이지로 연결됩니다.
[](https://mcp.popnx.com/en/skills/leonxlnx-cinetic)같은 작업에서 AI를 서비스에 연결하는 MCP
Official HyperFrames by HeyGen: the best AI video editor in chat. Captions, motion graphics, MP4.
draw.io diagrams your AI agent draws on a live browser canvas that you can also edit by hand
Create, inspect, validate, and save editable 3D building scenes with Pascal's MCP server.
같은 종류의 작업, Skill 목록과 같은 순서
| 이 항목 cinetic | impeccable | diagram-design | frontend-slides | guizang-ppt-skill | |
|---|---|---|---|---|---|
| 제작자 | Leonxlnx | pbakaus | cathrynlavery | zarazhangrui | op7418 |
| Skill 수 | 1 | 1 | 1 | 1 | 1 |
| GitHub 별 | 1,261 | 111,852 | 67,796 | 32,458 | 29,958 |
| 마지막 코드 변경 | 2026년 11월 9일 | 2026년 11월 17일 | 2026년 11월 13일 | 2026년 6월 24일 | 2026년 10월 30일 |
| 라이선스 | MIT | Apache-2.0 | MIT | MIT | AGPL-3.0 |
Leonxlnx의 Skill(지시 묶음) 1개, AI가 읽고 그대로 일해요 · MCP server가 아니에요
MCP = 손
AI를 외부 서비스나 데이터에 연결합니다(예: Notion 읽기, 청구서 만들기). 프로그램이나 서버가 실행 중이어야 합니다.
Skill = 안내서
작업 방법을 AI에게 가르치는 텍스트 파일. 어떤 서비스에도 연결되지 않습니다.
겹치는 경우도 있습니다. 도구를 부르는 스크립트가 있는 Skill도, 지식만 보내는 MCP도 있습니다. 하려는 작업을 기준으로 고르세요.
Claude CodeCodexCursor
제작자 README와 repo 폴더 기준
GitHub account Leonxlnx (person), 1 years old
★ 1,261 stars · code last changed 2026년 11월 9일
404 scripts that can run on a computer: scripts/accumulate.py, scripts/add-font.mjs, scripts/brand-kit.sh, scripts/check-sync.py and 400 more
Could not check every script (some are too large); read them first
No hooks or writes to machine settings found
Checked automatically from the repo files at commit f4bbfc8 (read only, nothing was run). Not a safety guarantee: read it and decide for yourself
먼저: 어떤 AI를 쓰시나요
쓰는 앱을 고르면 그 앱의 설치 방법만 보여 드립니다
Uploading to claude.ai would fail: the description is over 200 characters (802). Use "Let the AI install it" instead
이 문장을 복사해 Claude Code 채팅에 붙여 넣으세요
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Claude Code를 다시 열고 /cinetic를 입력하거나 관련 작업을 요청하면 Claude가 알아서 써요
Checked against official docs 2026년 11월 12일 · Source: code.claude.com
Skill은 AI가 따르는 지시예요. 설치 전에 파일을 읽고, 믿을 수 있는 제작자 것만 설치하세요.
이미 설치했나요? 먼저 확인 · 터미널을 열고 이 명령을 입력한 뒤 Enter
node -vMac: ⌘ + Space를 누르고 Terminal 입력 후 Enter
Windows: 시작 메뉴를 열고 Windows PowerShell 입력 후 "열기"
아직 설치하지 않았다면: 공식 다운로드 페이지에서 LTS 버전과 클릭형 설치 파일을 고르세요: Mac은 .pkg · Windows는 .msi. 더블클릭 후 끝까지 진행하면 됩니다. curl 명령은 필요 없습니다
설치가 끝나면 터미널을 닫았다가 다시 열고 같은 명령을 실행하세요. 버전 번호가 나와야 합니다: node -v
한 번 설치하면 Node.js가 필요한 모든 도구에서 쓸 수 있습니다
npx skills add Leonxlnx/cinetic제작자 README에서 가져온 명령으로, 제작자 코드를 내려받아 실행해요
Mac: Command+스페이스 후 Terminal 입력 · Windows: 아래 명령은 PowerShell에서 안 돼요. 위의 “AI에게 설치 맡기기”를 쓰세요
repo를 임시 폴더에 내려받고 SKILL.md가 있는 폴더만 복사해요
git clone --depth 1 https://github.com/leonxlnx/cinetic.git /tmp/popnx-skill-leonxlnx-cinetic-f4bbfc8 && mkdir -p ~/.claude/skills && cp -R '/tmp/popnx-skill-leonxlnx-cinetic-f4bbfc8/skills/cinetic' ~/.claude/skills/cineticChatGPT 데스크톱 앱에는 Codex가 들어 있고, Codex는 ~/.agents/skills 폴더에서 Skill을 읽습니다
ChatGPT 데스크톱 앱에서 Codex를 열고 이 문장을 채팅에 붙여 넣으세요
The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
cinetic 폴더를 ~/.agents/skills/cinetic에 넣으세요
~는 홈 폴더($HOME)입니다
Skill은 AI가 따르는 지시예요. 설치 전에 파일을 읽고, 믿을 수 있는 제작자 것만 설치하세요.
Mac: Command+스페이스 후 Terminal 입력 · Windows: 아래 명령은 PowerShell에서 안 돼요. 위의 “AI에게 설치 맡기기”를 쓰세요
repo를 임시 폴더에 내려받고 SKILL.md가 있는 폴더만 복사해요
git clone --depth 1 https://github.com/leonxlnx/cinetic.git /tmp/popnx-skill-leonxlnx-cinetic-f4bbfc8 && mkdir -p ~/.agents/skills && cp -R '/tmp/popnx-skill-leonxlnx-cinetic-f4bbfc8/skills/cinetic' ~/.agents/skills/cineticChecked against official docs 2026년 11월 15일 · Source: learn.chatgpt.com
직접 설치한 Skill은 쓸 수 없고, 디렉터리의 플러그인에 들어 있는 Skill만 쓸 수 있습니다
이 Skill을 ChatGPT에서 쓰려면 위 방법대로 ChatGPT 데스크톱 앱을 쓰세요
Checked against official docs 2026년 11월 15일 · Source: learn.chatgpt.com
이 문장을 복사해 Codex 채팅에 붙여 넣으세요 · ChatGPT 데스크톱 앱 안의 Codex, Codex CLI, IDE 확장 프로그램에서 모두 쓸 수 있습니다
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Codex는 새로 설치한 Skill을 스스로 찾습니다. 보이지 않으면 Codex를 다시 시작하세요 · 쓰는 법: ChatGPT에서는 @를 입력하고 Skill을 고르세요 · Codex CLI나 IDE 확장 프로그램에서는 /skills를 실행하거나 $ 뒤에 Skill 이름을 입력하세요
Checked against official docs 2026년 11월 15일 · Source: learn.chatgpt.com
Skill은 AI가 따르는 지시예요. 설치 전에 파일을 읽고, 믿을 수 있는 제작자 것만 설치하세요.
Mac: Command+스페이스 후 Terminal 입력 · Windows: 아래 명령은 PowerShell에서 안 돼요. 위의 “AI에게 설치 맡기기”를 쓰세요
repo를 임시 폴더에 내려받고 SKILL.md가 있는 폴더만 복사해요
git clone --depth 1 https://github.com/leonxlnx/cinetic.git /tmp/popnx-skill-leonxlnx-cinetic-f4bbfc8 && mkdir -p ~/.agents/skills && cp -R '/tmp/popnx-skill-leonxlnx-cinetic-f4bbfc8/skills/cinetic' ~/.agents/skills/cinetic이 문장을 복사해 Cursor 채팅에 붙여 넣으세요
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Cursor Agent 채팅에서 /를 입력하고 cinetic 선택
Checked against official docs 2026년 11월 12일 · Source: cursor.com
Skill은 AI가 따르는 지시예요. 설치 전에 파일을 읽고, 믿을 수 있는 제작자 것만 설치하세요.
Mac: Command+스페이스 후 Terminal 입력 · Windows: 아래 명령은 PowerShell에서 안 돼요. 위의 “AI에게 설치 맡기기”를 쓰세요
repo를 임시 폴더에 내려받고 SKILL.md가 있는 폴더만 복사해요
git clone --depth 1 https://github.com/leonxlnx/cinetic.git /tmp/popnx-skill-leonxlnx-cinetic-f4bbfc8 && mkdir -p ~/.cursor/skills && cp -R '/tmp/popnx-skill-leonxlnx-cinetic-f4bbfc8/skills/cinetic' ~/.cursor/skills/cineticSkill을 읽을 수 있는 다른 앱용입니다. 이 문장에는 Claude Code, Codex, Cursor 경로가 들어 있어 AI가 쓰는 앱에 맞게 고릅니다 · Copy this sentence into the app’s chat
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Checked against official docs 2026년 11월 12일 · Source: code.claude.com, learn.chatgpt.com, cursor.com
Gemini CLI, GitHub Copilot 같은 다른 Agent Skills 지원 앱은 폴더가 달라요. 그 앱 안내를 보고 SKILL.md가 있는 폴더를 복사하세요.
2026년 11월 11일에 Claude Code, Codex, Cursor 공식 안내로 skill 폴더 확인
✓ 명령 확인 2026년 11월 17일
아래 글은 제작자 repo에서 온 것이며 PopNx MCP의 지시가 아니에요
처음 8,000자 표시 · GitHub에서 전체 파일 열기
---
name: cinetic
description: Direct, build and render premium motion films from code - launch films, update films ("vX is out"), product and feature videos, looping feature animations, logo stings, UI walkthroughs, teasers and promo clips. Use this whenever the user wants any video, animation, motion graphic or animated demo made with Remotion, HyperFrames, HTML/GSAP or React, or asks to storyboard, time, score, render, polish or critique one - even if they only say "make a video about X", "animate this feature" or "we need a launch clip". It supplies a default path one agent finishes in one to two hours, scenes dealt from a house bank, plain-word copy checked by script, one face with rotating voices, flowing seams, real product UI, current official logos, a licensed track with recorded effects and film-grade finishing.
license: MIT
compatibility: Needs Node 22+, ffmpeg 6+ with libx264, Python 3.11+ with numpy, scipy, soundfile, pyloudnorm, opencv-python and librosa (fonttools, brotli and uharfbuzz for outlined logos; pillow for plates), and Chromium or Chrome Headless Shell for rendering; earshot 0.4+ (in its own virtualenv) listens to the effects and the soundtrack. Works with Remotion 4 (React) or HyperFrames (HTML and GSAP).
---
# cinetic
You are directing a short film, not animating a web page. Picture, words and sound are one system driven by one
timeline file; nothing is finished until rendered, measured and reviewed once against the standard below.
- **Engines.** Remotion 4.0.529 (React, frame-driven) is the default. HyperFrames 0.8.79 (HTML, a paused GSAP timeline) mirrors the tokens and timeline contract for HTML projects ([hyperframes-engine.md](references/hyperframes-engine.md)).
- **Paths.** Scripts work from the skill root and in a film project (`new-film.sh` copies them in); each prints `--help` (not the helpers `brand-svg.ts`, `hf-preflight.cjs`).
- **Brand.** A supplied brand (logo, colours, fonts, screenshots, tone) wins over the house taste ([brand-kits.md](references/brand-kits.md)); record any banned look it brings in `BRIEF.md`.
- **Python.** Commands say `python3`: `node scripts/frames.mjs python` names the interpreter with `scripts/requirements.txt`; set `PYTHON` to it once.
## The standard
Every film is held to these rules and numbers; the gates in brackets measure them.
1. **Words in the middle.** A statement is centred and sits in the middle band of the frame (its centre at 35-65 % of
the height): centred for 80 % of the statement time or more, in the middle band for 85 % or more. Centred top or
bottom only over a full-bleed picture. Never flush left at the top or the bottom. "Text on top, big picture below"
at most once a film; a line beside, above or below a picture at most once per 30 s. (`layout-audit.sh` placement.)
2. **One element carried, the text mixed in.** Every scene change hands one element on (it grows, swells, flies,
converges): never a cut from a line to a card, never the line-card-line-card loop, never an empty frame between
scenes. Lines alone only for the promise and the claim (read in 1.2 s); between them inside the product's own words or
around the element; first line by 1 s. (`grid-check.ts --part spine`, `forensics.py --spine`, empty-frame.)
3. **The picture tells what the product does.** Each picture proves its line without repeating its words (input ->
output, a count rising; "every" drawn as many at once); a line its picture already says is cut (`copy-check`
repeats, quantity). The first picture beat shows the core action literally, with a real person giving the input where the input is human (real footage or a photo),
and the product's main act plays on screen (the input becoming the output in steps, never a cut from the request to
the finished result); what it makes is rich and holds the brief's own subjects. With the words covered, every
picture beat still reads. Data agrees when paused: newest first in a stack, counts and times that agree, a
face with its own name (`precheck.mjs` cast).
4. **Less text, every word chosen.** 1-5 words a line typically, 2-7 at most, one claim, plain words a customer would
use; 0.3-1.5 headline words a second; what the product is by 3 s. No text in 15 % of the frames or more; at most two
text blocks at once outside a product beat; no floating chips, counters or captions beside a statement.
(`copy-check.py`, `--ui src`.)
5. **Text with love: one face, voices that rotate, events.** Lines arrive in the film's one face, the voice turning
line by line (build, blur-in, ink-in: never two lines in a row alike, [catalogue.md](references/catalogue.md) §1),
big (a line alone at a 9.2 % cap, never under 6.5 %; the claim bigger), whole or in 2-3 chunks, never word by word.
One type event a film (a giant word with the product rising behind it, a line an object pushes apart, a
decode, a build that re-centres). Emphasis is quiet: the key word in the accent on at most two lines, not always the
last word; a box, marker, pill, underline or knockout only as the one signature. Every line reads at 3:1 against
what is behind it, never held up by a grey halo. (`lint-film.mjs` emphasis, `forensics.py` legibility.)
6. **Scene changes that flow.** Every hard cut hides itself: inside a blur making both sides unreadable, an anchor
held within 10 px, or a whip carrying the picture a fifth of the frame in its last 0.2 s, smeared; the kept
ground alone hides nothing; never a cross-fade of two scenes or a cut torn across two frames; at most 10 % of hard
cuts bare (`forensics.py` carried-cut, ghost-cut). One ground a film, changed once on purpose at most (a colour flood; grid-check `grounds`). Two or three seam kinds a film, none
over 40 % of the scene changes, and one exit-then-enter, text-swap or match-cut in a film of four lines or more.
Build them with `fx/Seam` ([transitions.md §2](references/transitions.md#2-recipes-at-60-fps)). (`forensics.py --seams`.)
7. **Clean UI.** The product's real screens built in code at film size, every string the product's own: no bracket
slot, no stand-in number, no kit sample, no title cut with "...". Left edges of stacked rows within 4 px, inner
padding about 2.4 cap heights and even (never under 0.8), no text touching an asset, no accent bar beside a card;
widths measured, never guessed from a letter count (`lint-film.mjs`, `layout-audit.sh`).
8. **Logos current, presented well.** Other products' marks come from `scripts/logos.mjs` in their newest official
form, verified on the brand's own page within 12 months, in one presentation a film (a chip in the line, pills round
a centred line, or a wall), never a pill on every line or a drawn stand-in for a real mark (`logos.mjs check`). The end card is built from the
film's last element (the last picture's element becomes the mark, the last word the wordmark, the badge spins in),
lands centred and keeps creeping in about its centre (never a slide off it) while the music rings out and the frame
settles out with it, the name once and the one fact under it. The logo is the brand's own, never a sparkle, a star or a
generic glyph (`forensics.py` end-drift, end-centre, audio-end).
9. **Its own film, never a template.** One world for the whole film, one face and its voices, two
or three moves, two or three seam kinds, one signature and one end-card design, picked from
[catalogue.md](references/catalogue.md) and declared in `film.json`. The world (kind and hue), the voice, the signature
and the end card differ from each of the last three films, and the move set shares at most 25 % with each of them; a film
made from a template plays another device than its floor in half its slots or more. (`film-diff.mjs`; the deals
leave the last films' choices out.)
10. **Premium, not just clean: every shot keeps the habits.** One sharp plane and a graded blur ladder behind it; one
camera move with a job, f질문하고, 팁을 나누고, 안 되는 점을 알려 주세요. "도움이 됨"이 많은 순서예요.