Adya Video Skill
rashigupta-Adya/adya-video-skill
Gera vídeos on-brand de produção em tempo real via Claude Code
Ver no GitHubSkill para Claude Code que transforma tópicos em vídeos de marketing on-brand (launch, explainer, demo, talking-head) com script, voiceover via ElevenLabs, animação e subtítulos. Oferece dois motores de animação — HyperFrames (rápido, nativo para áudio/legendas) e Remotion (custom, física, 3D) — permitindo design único para cada vídeo.
Instalação
npx convez-ai rashigupta-Adya/adya-video-skill1 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
- Criar vídeos de lançamento de produto com animação cinemática
- Gerar vídeos explicativos para conceitos e features
- Produzir demos de UI com footage real e b-rolls editados
- Fazer talking-heads com voiceover automático e legendas
Veredito

Ferramenta robusta para equipes que precisam produzir vídeos de marketing em escala sem infraestrutura pesada. Exige investimento em setup (Node, ffmpeg, chaves de API), mas democratiza a criação de conteúdo profissional dentro do Claude Code. Custo marginal baixo (ElevenLabs + Higgsfield opcionais).
README
A tradução para o português deste README ainda está na fila.
adya-video — installable skill
Turn a topic into an on-brand Adya marketing / explainer / demo video: script → ElevenLabs voiceover → animation → MP4, with a review loop and optional Higgsfield AI visuals (pay-as-you-go). Built so anyone on the team can run it from Claude Code. Two first-class animation engines — HyperFrames (fast, audio/caption-native) and Remotion (custom/physics/3D, reusable components) — chosen per video or combined.
What's here
adya-video/
├── SKILL.md # the orchestrator workflow (start here)
├── assets/adya-logo.webp # the ONLY constant across videos (placed in every one)
├── scripts/elevenlabs_tts.sh# voiceover generation + duration measurement
└── references/ # setup, variety (anti-sameness), video-types, design-system, engines,
# script-style, hyperframes-build, music, product-explainer,
# higgsfield, remotion, toolbox
No house palette, fonts, or template. Every video gets a unique design (user-specified or invented). The Adya logo is the only thing every video shares.
What it can make
Four video types:
- Launch — heavily animated, cinematic, imagery-led ("make it sexy"). Announcements, teasers, homepage hero.
- Explainer — concept-driven motion graphics. "What is X / how it works / the thesis."
- Demo — real UI screens + features. Team-shot footage is edited first (
video-use), then b-rolls/callouts/captions layered on. - Tutorial / talking-head — a real person on camera + b-rolls + the face-to-corner circle + captions; keeps their real voice.
Across every type: script in Adya's voice (grounded in real product facts), human voiceover (ElevenLabs) or kept real voice or free local TTS, optional background music and optional subtitles (you choose per video), b-rolls, optional AI visuals (Higgsfield / FLUX), diagrams/infographics, unique design every time, a replicate-or-fresh variety control, and any aspect ratio (landscape / vertical / square). Two engines — HyperFrames + Remotion — chosen per video or combined.
Install (per teammate)
Option A — from git (recommended): in Claude Code or a terminal:
npx skills add rashigupta-Adya/adya-video-skill
The skills CLI installs adya-video/SKILL.md + its bundled references into ~/.claude/skills/adya-video/. Restart your agent session after installing.
Option B — manual: download this repo, copy the adya-video/ folder into ~/.claude/skills/, then restart the agent session.
After install, the skill triggers when you say things like "make a 15s Adya video about AGP" or "turn the Recruitment Agent into an explainer." On first run it checks your environment and offers to install anything missing (see adya-video/references/setup.md).
One-time prerequisites (the skill checks these for you)
- Node 18+, ffmpeg, Chrome (HyperFrames manages it).
- HyperFrames skills:
npx skills add heygen-com/hyperframes. - Your own ElevenLabs API key →
export ELEVENLABS_API_KEY=sk_...(each person brings their own). - Optional (only for AI visuals): Higgsfield CLI +
higgsfield auth login(your own credits).
Cost model
Free by default (HyperFrames graphics + local render). The only paid steps are ElevenLabs voiceover (cheap; free tier ~10k chars/mo) and Higgsfield AI visuals (opt-in, gated, image-over-video). Keys and billing are per-person.