Communitygithub.com

editorial-and-reframe

Video and animation skill. Reframes horizontal 16:9 footage into vertical 9:16 mobile formats using 1-Euro adaptive subject tracking, collision-safe slot lanes, and title-safe boundary enforcement. Use when adapting widescreen videos, reframing talking heads, or positioning graphic overlays.

O que é editorial-and-reframe?

editorial-and-reframe is a Claude Code agent skill that video and animation skill. Reframes horizontal 16:9 footage into vertical 9:16 mobile formats using 1-Euro adaptive subject tracking, collision-safe slot lanes, and title-safe boundary enforcement. Use when adapting widescreen videos, reframing talking heads, or positioning graphic overlays.

Funciona com~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/Vichith-AI/vichith-motion/tree/HEAD/skills/editorial-and-reframe

Perguntar na sua IA favorita

Abre um novo chat com esta habilidade de agente já pré-carregada.

Documentação

Editorial and Reframe

Intelligently adapt landscape (16:9) video footage into vertical (9:16) compositions while keeping active subjects perfectly framed and graphics free from overlap.

1. 1-Euro Adaptive Subject Tracking

Naive camera framing on raw bounding boxes causes severe jitter during slow movements and lag during fast turns. The tracking engine uses a 1-Euro adaptive low-pass filter:

  • Filter Tuning:
    • minCutoff = 1.0: Suppresses high-frequency micro-shakes when the speaker is resting.
    • beta = 0.007: Dynamically reduces filtering latency when velocity spikes, preventing tracking lag.
    • dCutoff = 1.0: Derivative filter cutoff for reliable velocity estimation.
  • Anchor Offsets: Layers can lock to dynamic subject landmarks:
    • above_head: Floating name tags and speaker labels ($y_{\text{top}} - 0.05$).
    • flank_left / flank_right: Automatically detects the open space opposite the speaker's face direction.

2. Collision-Safe Slot Lanes

When multiple text blocks, badges, or lower-thirds target the same visual sector, slot lanes prevent collisions through deterministic stacking:

  • Top Lane: Stacks downward (top_center hero remains at anchor; subheadings push downwards with a 0.018 gap).
  • Bottom Lane: Stacks upward (bottom_center captions stay pinned to safe baseline; tags push upwards).
  • Safe Zone Boundary: A mandatory 0.06 margin (6% viewport inset) protects text from being cropped by mobile app chrome or rounded screen corners.

3. Reframe Workflow

  1. Analyze Video Trajectory: Ingest subject bounding boxes across the timeline.
  2. Smooth Trajectory: Pass coordinates through OneEuroFilter to generate smooth pan-and-scan focal points.
  3. Calculate Crop Window: Center the 9:16 window on the smoothed coordinate while clamping within the source 16:9 frame.
  4. Assign Graphic Slots: Place captions, watermarks, and callouts into distinct semantic slots (top_center, bottom_center).

4. Declarative MotionSpec Example

{
  "$schema": "https://vichith.in/schemas/motion-spec.json",
  "reframe": {
    "sourceAspect": "16:9",
    "targetAspect": "9:16",
    "subjectTracking": {
      "target": "speaker_face",
      "filter": "1-euro",
      "params": { "minCutoff": 1.0, "beta": 0.007 },
      "leadRoom": 0.04
    },
    "slots": [
      {
        "id": "speaker-badge",
        "slot": "top_center",
        "height": 0.06,
        "priority": 0
      },
      {
        "id": "episode-tag",
        "slot": "top_center",
        "height": 0.035,
        "priority": 1
      },
      {
        "id": "synced-captions",
        "slot": "bottom_center",
        "height": 0.08,
        "priority": 0
      }
    ]
  }
}

Individual skills in this repo

This repo contains 2 individual skills — each has its own dedicated page.

Habilidades Relacionadas