Skip to main content
skill CX Codex

key-layout — Claude Code skill for video key layouts

key-layout is a Claude Code skill that places a video's main topic at the top and related points above the captions. It activates when keyLayoutEnabled is on in the brief or the edit prompt includes “Bố cục Key: BẬT”.

MIT ★ 147 updated 2026-11-13

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add notivn/AIEVv2 --skill key-layout -a claude-code
Open on GitHub

What key-layout does

key-layout defines where the main topic and related text cues appear in a video. The main cue stays in the top band; related cues appear above the captions at the timestamps where each point is spoken. When the user supplies no wording, the skill derives the main phrase from the transcript and forms related cues from the video's individual points.

The bands stay fixed during zooms and must not collide with captions, cards, or TikTok safe zones. Verification requires extracting frames with ffmpeg and reading them with Read. A band placed inside #face-wrapper moves with the zoom and drifts out of position.

What it helps with

  • place the video's main topic at the top of the frame
  • show related points at their spoken timestamps
  • keep key bands clear of captions and cards
  • check text legibility on bright footage

Example prompts for the agent

  • Put the video's main topic at the top and show related points above the captions at their spoken timestamps.
  • Turn on the key layout with keyLayoutEnabled.
  • Apply the key layout from the “Bố cục Key: BẬT” section of the edit prompt.

How to install key-layout

1Skills CLI (Vercel)

npx skills add notivn/AIEVv2 --skill key-layout -a claude-code

2Manually: put the files in place

~/.claude/skills/key-layout/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/key-layout/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about key-layout

How do I install key-layout in Claude Code?

Run npx skills add notivn/AIEVv2 --skill key-layout -a claude-code. The skill file is at ~/.claude/skills/key-layout/SKILL.md.

What does key-layout do?

It places the video's main topic in a top band and related points above the captions. Each related cue appears when its point is spoken.

What does key-layout use to verify the video?

It requires extracting frames with ffmpeg and reading the images with Read to check placement, legibility, and behavior during zooms.

Source text

SKILL.md · MIT
---
name: key-layout
description: Main key / related key layout on the video - the MAIN KEY (topic/hook) sits in a band at the TOP of the video, the RELATED KEYS in a band at the BOTTOM (above the caption area), timed to whatever is being said. Read this when the brief enables "Bố cục Key" (keyLayoutEnabled) or the edit prompt contains the section "Bố cục Key: BẬT".
---

# Key Layout - main key on top, related keys below

Goal: a viewer scrolling past instantly knows **what the video is about** (main key, always on top)
and **which point is being made right now** (related key, changes with the content, at the bottom).

## ⚖️ STYLE DESIGN - PRECEDENCE RULE

If the edit prompt has a "STYLE DESIGN (BẮT BUỘC TUÂN THỦ 100%)" section -> the key colors/fonts/tone come
ENTIRELY from that style. If it does not -> use the default branding of the format skill in use.

## Choosing the keys (when the user does not specify them)

- **MAIN KEY**: ONE phrase of 2–6 words summarizing the topic/hook of the whole video (e.g. "MCP chính thức của TikTok",
  "Review iPhone 17 Pro"). Take it from the transcript - prefer a phrase that is repeated often or sits in the hook sentence.
  Do NOT use a whole long sentence; no trailing punctuation.
- **RELATED KEYS**: 3–6 phrases of 1–4 words, each one a SINGLE point made in the video (a feature, a number,
  a proper noun, a process step). Each key is pinned to the timestamp where that point is mentioned (from the transcript's
  word timestamps). If the user supplies a list -> use ALL of it, in the order the content mentions them.

## Band positions (9:16 - 1080×1920; other ratios convert by %)

| Band | y range | Notes |
|---|---|---|
| **MAIN KEY (top)** | ~96–345px (5–18% of height) | Horizontally centered; AVOID covering the face (faces usually start ~20%) |
| **RELATED KEYS (bottom)** | ~1290–1490px (67–78%) | ABOVE the caption area (captions sit at `bottom: ~372px` = y>=1548) |

- 16:9 (1920×1080): main key at y ~54–190px; related keys at y ~700–840px, captions at the very bottom.
- The TikTok safe zone still applies: bottom 15% + right 12% - the bottom band must not spill into those two areas.
- Both bands are STATIC layers relative to the camera: **do NOT put them inside the zoomed wrapper** (`#face-wrapper`) -
  just like captions, otherwise the keys drift/get clipped on punch-in.

## Typography & style

- **Main key**: the style's heading font, weight 800, size ~64–88px (at 1080w), 1 line (if too long, reduce the
  size; do not wrap to 2 lines unless <=2 words per line). Primary color or a primary→secondary gradient
  (`background-clip:text` - you MUST include the Vietnamese diacritics fix, production-verified in the
  noti-tiktok-vn skill: `line-height: 1.0` + `padding-top: 0.5em` + `padding-bottom: 0.14em`).
- **Related keys**: pill/chip - font size ~38–48px weight 700, padding `14px 28px`, 999px radius,
  glass background (`rgba` of the style background + blur) or a soft accent background, text in the style's text/accent color.
- **Readable over any footage**: behind the main key band add a soft scrim
  (`background: linear-gradient(180deg, rgba(0,0,0,0.45), transparent)` covering ~0–20% of the top of the video,
  color taken from the style's background) or a strong text-shadow - verify with a snapshot on the BRIGHTEST frame.

## Timing & animation (GSAP - deterministic, no setTimeout/random)

- **Main key**: enters ONCE at ~0.3–0.8s (reveal y 30→0 + opacity, `power3.out` 0.5s), then
  STAYS PUT for the whole video (a very subtle glow/breathing scale of ±1.5% with a cycle >=3s is allowed). For a long video split
  into chapters, the main key may change once per chapter.
- **Related keys**: each key enters exactly when its point starts being said (per the word timestamps), holds 2.5–4s
  then exits. Enter: y 24→0 + opacity 0→1, 0.35s `power3.out`; exit: opacity→0 + y→-12, 0.3s. Only 1 key
  visible at a time (up to 2 if two points run together); keys are >=1s apart.
- A related key appearing SHOULD line up with a light accent SFX (`pop.mp3`/`ting`, volume 0.4) if the brief enables SFX.

## Working with the other layers

- Main key ↔ kinetic typography: when a scene has a large kinetic title in the top area (e.g. the hook scene),
  HIDE the main key during that scene (avoid two large texts on top of each other) - the main key returns from the next scene.
- Related keys ↔ captions: two separate layers that must never overlap (the bottom band sits ABOVE the captions).
- Related keys ↔ spec cards/stat callouts: while a card occupies the bottom band, DELAY the key until the card exits.

## Verify (mandatory before reporting done)

Extract frames with ffmpeg and `Read` each image:
1. Frame at ~1s: the main key has entered, sits in the top band, does not cover the face, and its Vietnamese diacritics are complete (zoom 3×).
2. A frame in the middle of each related key: correct bottom band, not covering the captions, not spilling into the safe zone.
3. A frame during a punch-in zoom: both bands STAY PUT (they must not zoom along).
4. The brightest frame of the footage: the keys are still clearly readable (scrim/shadow is sufficient).

## Known issues

- Vietnamese gradient text losing its diacritics -> apply the `line-height: 1.0` + `padding-top: 0.5em` + `padding-bottom: 0.14em` fix (verified, see noti-tiktok-vn).
- Putting a band inside `#face-wrapper` -> the key scales with the zoom and drifts out of place. The band must be a sibling.
- The main key covering the hook title in the opening scene -> hide the main key in scenes with a large kinetic title.
- Too many related keys flashing in a row -> viewers get tired; keep the pace at >=1 key per 4s, each key >=2.5s.

© notivn · MIT · notivn/AIEVv2

Similar skills

All skills for Design →

From the same set15

The whole set →
skill-authoring skill-authoring is a Claude Code skill that sets the rules for creating and revising skills for AI Edit Video. It activates on requests to write a skill or capture a lesson after a video is finished. CX Codex +1 ★ 147repo stars color-grading color-grading is a Claude Code skill that handles video color grading in the AI Edit Video pipeline. Runs when processing HDR or log clips, or upon a Chỉnh màu request. CX Codex +1 ★ 147repo stars background-music background-music is a Claude Code skill that selects a soundtrack and configures automatic ducking under speech in Remotion videos. It runs when a brief requests background music or an edit asks to add or adjust it. CX Codex +1 ★ 147repo stars offline-voice offline-voice is a Claude Code skill that guides the setup of VieNeu-TTS narration and voice cloning. Activates when working on /api/tts, /api/voices, or narration debugging. CX Codex +1 ★ 147repo stars make-a-video make-a-video is a Claude Code skill that turns a concept into a finished MP4 with HyperFrames. It gathers requirements, builds a storyboard, and checks the preview before the final render. CX Codex +1 ★ 147repo stars auto-cut auto-cut is a Claude Code skill that cuts silences, stutters, and repeated takes from talking-head footage before editing. It activates when the autoCut brief setting is enabled. CX Codex +1 ★ 147repo stars ai-illustrations ai-illustrations is a Claude Code skill that adds generated images to a video edit. It activates when the brief enables AI illustrations or requests them for a video. CX Codex +1 ★ 147repo stars remotion-assemble remotion-assemble is a Claude Code skill that stitches HyperFrames scenes, footage, voiceovers, and SFX into a video via meta.json. It activates when editing files inside engines/remotion. CX Codex +1 ★ 147repo stars

Project sponsors

Companies that keep this analytics open