VEED Skills
veedana/veed-skills
Skills para criar vídeos com lip-sync, subtítulos e remoção de fundo via agentes IA
Ver no GitHubConjunto de skills para geração de vídeos com IA, incluindo lip-sync, subtítulos estilizados e remoção de fundo. Funciona com Claude Code, Cursor e qualquer agente que suporte skills em Markdown, integrando com APIs da VEED na plataforma Fal.
Instalação
npx convez-ai veedana/veed-skills1 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
- Gerar vídeos de apresentadores com lip-sync a partir de imagens estáticas
- Criar vídeos de pitch de produtos com apresentador sincronia labial
- Adicionar subtítulos estilizados e queimados em vídeos com 27 presets visuais
- Remover fundos de vídeos com controle de foreground refinement
Veredito

Skills bem estruturadas e modularizadas para automação profissional de vídeo com IA. Exige API key (Fal) e integração com agentes específicos, mas oferece pipelines prontos para casos de uso complexos em produção de conteúdo visual.
README
A tradução para o português deste README ainda está na fila.
VEED Skills

Agentic AI Video Creation
Drop-in skills that teach your AI agent to make video with VEED — talking heads, product pitch videos, styled subtitles, and background removal — powered by VEED's APIs on Fal.
Works with Claude Code, Codex, Cursor, and any AI agent that supports Markdown-based skills.
"Make this photo say 'Welcome to our store' in a confident British accent.
Then remove the background and add animated subtitles in the glass preset."
→ generates the talking-head video → strips the background → burns captions → returns three URLs
What's included
Five skills — four endpoint skills and one workflow skill — that work standalone or chain together:
| Skill | What it does | Invoke |
|---|---|---|
| veed-talking-head | Static image + audio file → lip-synced presenter video. Powered by VEED's Fabric 1.0. | /veed:talking-head |
| veed-talking-head-text | Static image + text script → lip-synced video. AI voice generation built in (optional voice description). | /veed:talking-head-text |
| veed-subtitles | Any video → styled, burned-in subtitles. 27 visual presets, auto-transcription in 125+ languages, full positioning control. | /veed:subtitles |
| veed-background-removal | Any video → clean subject. Three modes: standard, fast, green screen. Optional foreground refinement. | /veed:background-removal |
| veed-product-pitch | Product image/description + spokesperson image/description + script/audio → finished product spokesperson video with optional subtitles. Chains image generation, talking head, and subtitles in one flow. | /veed:product-pitch |
Install
The repo ships five skills you can install independently. Pick whichever subset you need.
Option 1 — gh skill install (Claude Code, Cursor, Codex, Gemini CLI, Copilot, and more)
gh skill install veedana/veed-skills veed-talking-head
gh skill install veedana/veed-skills veed-talking-head-text
gh skill install veedana/veed-skills veed-subtitles
gh skill install veedana/veed-skills veed-background-removal
gh skill install veedana/veed-skills veed-product-pitch
Requires GitHub CLI v2.90+. The CLI writes to the right directory for your agent automatically. See INSTALL.md for per-host paths and full options.
Option 2 — Git clone
git clone https://github.com/veedana/veed-skills.git ~/.claude/skills/veed-skills
After cloning, all five skills are auto-discovered at veed-talking-head/SKILL.md, veed-talking-head-text/SKILL.md, veed-subtitles/SKILL.md, veed-background-removal/SKILL.md, and veed-product-pitch/SKILL.md.
Authentication
Two paths, in order of preference: MCP (preferred) → direct API with FAL_KEY.
| Priority | Mode | Trigger | Best for |
|---|---|---|---|
| 1 | Fal MCP (Bearer auth) | mcp__fal-ai__* tools visible in toolset | Claude Code, Cursor, Codex |
| 2 | Direct API (fal_client / curl) | FAL_KEY env var set | Any agent that can shell out or run Python |
Both use the same FAL_KEY from https://fal.ai/dashboard/keys. You're billed per model run at standard Fal rates either way — the MCP server itself is free.
Connect Fal MCP (Claude Code example)
claude mcp add --transport http fal-ai https://mcp.fal.ai/mcp \
--header "Authorization: Bearer $FAL_KEY"
Caveat: Fal's MCP currently supports API-key Bearer auth only. OAuth is "coming soon" per Fal — until it ships,
claude.aiCustom Connectors and Claude Desktop (which require OAuth) can't reach it. Claude Code, Cursor, Codex, Windsurf are unaffected.
Direct API fallback
export FAL_KEY=your_key_here
The skills detect FAL_KEY and use fal_client.subscribe(...) or curl against fal.run/veed/* endpoints.
Things to try
Once installed, try these prompts:
| Prompt | What happens |
|---|---|
| "Use veed-talking-head with this image and this audio file" | Animates the face, lip-syncs to the audio, returns the video URL |
| "Use veed-talking-head-text to make this photo say 'Our biggest launch yet' in a confident, warm voice" | Generates speech, animates, returns the video URL |
| "Use veed-subtitles to add animated captions to this video using the glass preset" | Auto-transcribes, styles, burns in captions, returns the video URL |
| "Use veed-background-removal on this clip in green-screen mode" | Removes the background, returns a clean subject video |
| "Take this product clip, remove the background, and add subtitles in the whisper preset" | Chains background-removal → subtitles, returns both URLs |
| "Create a product pitch video: here's my product [image], use a friendly female spokesperson, and have her say 'Check out our new launch'" | Generates a product composite, animates the spokesperson, optionally adds subtitles |
Pricing
Per-skill costs (charged by Fal — these skills add no markup):
| Skill | Cost |
|---|---|
| veed-talking-head / veed-talking-head-text | $0.08/sec (480p) · $0.15/sec (720p) · max 30s per generation |
| veed-subtitles | $0.10/min input video · 2× for >1080p · 2× for dynamic presets · 1-min minimum |
| veed-background-removal | $0.0225 per 30 frames (refine ON) · $0.012 per 30 frames (refine OFF) |
| veed-product-pitch | Composite of image generation + talking-head + optional subtitles costs (see individual skills above) |
Powered by VEED
VEED is the world's leading AI video platform. Learn more at https://veed.io/api.
License
MIT — Copyright (c) 2026 VEED LIMITED.