Motion Mosaic Editing
pouyA-png/motion-mosaic-editing
Pipeline agentic de motion graphics: decomponha, especifique, reutilize
Ver no GitHubSkill para edição agentic de motion graphics no estilo Mosaic, com metodologia de pipeline para camadas de gráficos em movimento, legendas cinéticas e infográficos sobre vídeos existentes. Fornece modelo mental de cinco princípios, workflow de seis etapas e template JSON reutilizável para especificação de cenas, agnóstico de engine (compatível com Remotion, SVG+Chromium, Canvas-WebGL e FFmpeg).
Instalação
npx convez-ai pouyA-png/motion-mosaic-editing1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Adicionar gráficos em movimento e legendas cinéticas a vídeos existentes
- Reformatar um vídeo para múltiplas plataformas (TikTok, Instagram, etc) mantendo editable scene spec
- Construir pipelines de edição repetíveis e inspecionáveis com IA
- A/B-testes de variantes de edição a partir de um único arquivo fonte
Veredito

Skill bem-documentado e metodologicamente sólido para edição de vídeo agentic, com ênfase em workflow e especificação declarativa. Excelente para quem quer orquestrar motion graphics com IA de forma replicável, mas exige familiaridade com pipelines técnicos e engines de render.
README
A tradução para o português deste README ainda está na fila.
motion-mosaic-editing
A Claude skill (Agent Skill) for Mosaic-style agentic AI motion-graphics video editing — a methodology for layering AI-generated motion graphics, kinetic captions, contextual overlays, and infographics onto footage you already have, orchestrated as a reusable, inspectable node/tile pipeline.
It is a methodology + workflow skill, distilled from research into mosaic.so and its "Motion" product. It teaches an agent how to plan and drive such an edit; it is engine-agnostic and ships no renderer of its own.
What it gives you
- A five-principle mental model (edit don't generate; decompose into Tiles; chains become reusable Agents; understand footage before editing; keep an editable scene spec).
- A six-step workflow: intake → transcribe & analyze → design the Tile pipeline → spec the motion-graphics brief → reframe/caption/score → export & reuse.
- A reusable JSON scene-spec template for element-addressable motion graphics (the technique that lets you tweak one title or value without re-rendering).
- An engine-agnostic implementation path: decide (LLM) → render (Remotion / SVG+Chromium / Canvas-WebGL) → encode (FFmpeg).
- Honest provenance: what is confirmed about Mosaic vs. inferred (see
references/research-provenance.md).
Install (Claude Code)
Clone into your skills directory:
git clone https://github.com/pouyA-png/motion-mosaic-editing.git \
~/.claude/skills/motion-mosaic-editing
Claude Code auto-discovers it. Invoke with /motion-mosaic-editing, or it
triggers on phrases like "add motion graphics", "animate this video",
"Mosaic-style editing", "auto-edit my footage", "kinetic captions",
"reframe for TikTok".
Structure
motion-mosaic-editing/
├── SKILL.md # entry point: principles + workflow
└── references/
├── architecture.md # Canvas, Tile catalog, Agents, triggers
├── motion-graphics-method.md # the brief method + JSON scene-spec
├── pipeline-stages.md # analysis, captions, reframe, models, export
└── research-provenance.md # confirmed vs. inferred + forensic findings
When to use
Adding motion graphics/overlays/infographics to a video; generating animated captions; reframing one video for multiple platforms; building a repeatable editing pipeline; A/B-testing edit variants from one source; or designing / critiquing an agentic-video-editing system.
License & disclaimer
MIT (see LICENSE). Methodology distilled from public research; not affiliated
with or endorsed by Mosaic, and no proprietary code or assets are reproduced.