immamdouhaboammar/motion-graphics-skills

jump-cut-pacing

Tighten raw talking-head footage with silence trimming and intentional face-safe jump cuts.

查看源码

使用方法

  1. 复制安装命令,在自己的智能体环境安装前先检查源码包。
  2. 从原始文档核对所需工具、凭据和支持的宿主。
  3. 说明输入材料与期望交付物,并对照来源检查输出。
npx skills add https://github.com/immamdouhaboammar/motion-graphics-skills --skill jump-cut-pacing

开始前需了解

  • 安装不包含模型订阅、API 额度或访问文件的权限。
  • 仓库星数和安装量反映活跃情况,不等于独立质量或安全测试。
仓库原始内容

按源仓库内容呈现,保留标题、案例、代码、表格、链接以及原文引用的演示图片。

Jump Cut and Pacing

Treat pacing as editorial direction, not automatic silence deletion.

The goal is to remove dead air while preserving breath, thought, emphasis, and the speaker's physical continuity. A fast cut is not automatically a better cut.

Decision gate

Use this skill when the source is mainly a presenter or interview and the editing problem is cadence.

Do not use it as the main solution when:

  • pauses are intentional dramatic beats
  • the source depends on performance timing, comedy, or music sync
  • a cut would remove a breath needed for intelligibility
  • the reference uses continuous camera movement rather than editorial punch-ins
  • the speaker already changes framing naturally between cameras

When uncertain, preserve the pause and flag it for review.

Editorial rules before automation

  1. Keep micro-pauses that make speech human.
  2. Cut retakes, abandoned starts, and dead air more aggressively than sentence breathing.
  3. Place cuts at phrase boundaries when possible.
  4. Avoid two hard cuts inside one short clause.
  5. Use a punch-in to clarify a cut, not on every sentence by habit.
  6. If the speaker moves strongly across a cut, prefer a wider frame or b-roll bridge instead of a tighter crop.

Read references/pacing-rules.md before changing the default thresholds.

Workflow

Step 1: Build the cut plan

Use the bundled planner:

bash
python3 scripts/cut_plan.py "$WORK/src.mov" --min-silence 0.45 -o "$WORK/cut_plan.json"

The planner probes the source duration with ffprobe when total duration is omitted. FFmpeg or ffprobe failure is a hard failure and must not produce a successful empty plan.

The plan contains:

  • source start and end for every kept speech segment
  • preserved lead padding before speech
  • preserved tail padding after speech
  • alternating zoom values
  • zoomanchory for face-safe vertical framing

Step 2: Review the plan as an editor

Before rendering, inspect:

  • whether a meaningful pause was classified as dead air
  • whether any kept segment is too short to read naturally
  • whether two cuts happen too close together
  • whether zoom alternation matches emphasis rather than becoming a metronome

If a section feels rushed, restore time. Do not lower the silence threshold globally just to fix one sentence.

Step 3: Render each segment with its planned framing

A cut plan is not complete until the zoom fields affect the pixels.

For every segment with source interval S to E, zoom Z, and anchor A, build a video branch using this pattern:

text
[0:v]
trim=start=S:end=E,
setpts=PTS-STARTPTS,
scale=iw*Z:ih*Z,
crop=iw/Z:ih/Z:(iw-ow)/2:(ih-oh)*A
[vN]

[0:a]
atrim=start=S:end=E,
asetpts=PTS-STARTPTS
[aN]

Then concatenate the matching video and audio branches in plan order:

text
[v0][a0][v1][a1]...[vN][aN]concat=n=SEGMENT_COUNT:v=1:a=1[outv][outa]

For Z equal to 1.0, this resolves to the original framing. For a punch-in, the scaled frame is cropped back to delivery size with the vertical crop controlled by zoomanchory.

If the active project already uses HyperFrames or another deterministic timeline, translate the same segment plan into that runtime instead of building a second renderer.

Step 4: Protect the face

Verify at least:

  • crown and hair are not clipped
  • eyes stay comfortably above center
  • chin is not trapped against captions
  • hand gestures are not removed by a punch-in
  • the crop still works if the presenter leans during the segment

Step 5: Protect audio continuity

Use the camera or production master audio. Do not reintroduce transcription audio into the final edit.

Check the assembled output for:

  • clipped consonants
  • missing word tails
  • clicks at cuts
  • sudden room-tone changes
  • accidental loudness pumping

Rhythm patterns

A useful sequence can be:

wide -> punch -> wide -> hold -> punch

A weak sequence is:

wide -> punch -> wide -> punch -> wide -> punch

The second pattern exposes the automation. Break the pattern when the meaning changes.

Creative QA

Hard-fail the edit if:

  • the speaker sounds unnaturally breathless
  • zooms happen with no semantic reason
  • face framing becomes less stable after the edit
  • the same two zoom states repeat for the whole video
  • a cut lands inside a meaningful word or gesture
  • the output is shorter but less comprehensible

What I learned the hard way

  • Centered crops often clip foreheads in vertical talking-head footage.
  • Removing every pause makes natural speech feel synthetic.
  • Alternating zoom is useful camouflage, but visible repetition becomes its own defect.
  • The rendered video is the evidence. A correct JSON plan is not proof that pacing works.
来自同一仓库

更多 Skills

全部 Skills
immamdouhaboammar
社区

animated-chart

Make a looping animated chart for LinkedIn or a newsletter in code with Claude Code. Bars that become a line, a number that counts up, a before and after. Every value stays exactly as given. Use when someone says "animate my chart", "animated graph", "make my results move", "stat animation" or wants a data graphic that loops.

1小时安装量
1
GitHub Stars
45
最近更新
11月1日
immamdouhaboammar
社区

apple-launch-film

Rebuild an Apple-style Mac launch film entirely in code with Claude Code. A thin menu bar, the black notch growing into widgets one scene at a time, full-screen gradient wallpapers and one short caption per scene. No screen recording, no After Effects. Use when someone says "Apple-style launch", "Mac launch video", "notch animation", "make it look like an Apple keynote", "macOS UI launch film" or wants a product launch that looks like native Mac interface.

1小时安装量
1
GitHub Stars
45
最近更新
11月1日
immamdouhaboammar
社区

assetseeker

Search free commercial-use creative assets across multiple online sources - photos, illustrations, icons, video footage, music, sound effects, and fonts. Use when the user needs to find assets for videos, PPTs, articles, or any content creation. Trigger on phrases like "find me a photo of", "search for icons", "I need background music", "find video footage", "search fonts", "找一张图片", "搜素材", "帮我找图标/视频/音乐/字体". Covers Pexels, Unsplash, Pixabay, Iconify, Freesound, Google Fonts and more - with API-ba

1小时安装量
1
GitHub Stars
45
最近更新
11月1日
immamdouhaboammar
社区

audio-mastering-mixer

Mix and master voiceover, sound effects, and music stems to professional broadcast loudness standards. Use when adding SFX cues, balancing audio stems, mastering to minus 14 LUFS, ducking music under voice, or eliminating harsh whooshes.

1小时安装量
1
GitHub Stars
45
最近更新
11月1日