convez.ai
SkillDesign & UIAvançado
clarkvalberg

Studio Video Creator

clarkvalberg/studio-video-creator

Gera vídeos de conceito profissionais direto do Claude com roteiro e render.

Ver no GitHub
InicialPluginScore · de 0 a 100

Uma skill Claude que produz vídeos conceituais de 60-90 segundos para produtos e projetos, combinando roteiro estruturado, design direction e render via Remotion. Inclui síntese automática de briefings, variantes de script, motion board interativo e voiceover com ElevenLabs.

Instalação

$npx convez-ai clarkvalberg/studio-video-creator

1 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

Estrelas
2
Forks
0
Acompanhando
0
Saúde
57%
Último commit
há 5 meses
Versão
v1.3.0
Issues abertas
0
Contribuidores
2

Casos de uso

  • Criar vídeos conceituais para apresentar produtos a investidores
  • Produzir filmes de protótipos para validação com times internas
  • Gerar ativos visuais para lançamentos de early-stage ventures

Veredito

Ferramenta robusta e bem arquitetada para produção de vídeos conceituais, com fluxo estruturado em 7 fases e gates de aprovação. Recomendada para times de produto e startups que precisam comunicar conceitos com clareza visual e narrativa.

README

A tradução para o português deste README ainda está na fila.

License: MIT Claude Skill Default Renderer: Remotion Optional Renderer: HyperFrames Voice: ElevenLabs Status: v1.3.0

A Claude skill that produces sixty-to-ninety-second concept films for product concepts, studio projects, and early-stage ventures.

Quickstart · How it works · Philosophy · Examples · Docs


What this is

A Claude skill, not an app. Drop it in, give it source material — a brief, a prototype, a deck, a URL — and tell Claude you want a concept film. The workflow produces a rendered 60–90 second video, with the first 15 seconds rendered first for fast review.

Studio Video Creator is built for short, voiceover-led concept films: prototype-grounded, insight-led, and designed to make an early product concept easier to understand. It fits the moment when a concept needs to be framed for a team, investor, partner, or launch audience.

It is closer to a thinking artifact than a marketing asset: structured enough to explain the concept, concrete enough to show what the product might become.

Why it exists

Concept films are won on recognition, not persuasion.

The viewer should understand the concept quickly enough to respond to the actual idea, not to the ambiguity around it. Many product videos drift into feature lists, generic claims, or visual polish that does not explain the product.

This skill keeps the film anchored in five principles:

  • The product is the protagonist's tool, not the protagonist's identity
  • The voiceover speaks to the viewer, not at them
  • Real screens, not abstract metaphors
  • The music builds; the film lands
  • Aspiration is earned through specificity, not asserted through superlatives

The goal is to make that workflow repeatable.

How it works

Seven numbered phases, plus a motion-board gate between script and design. Each one is a checkpoint.

#PhaseWhat happensGate
1IntakeClaude reads everything you provide and produces a two-sentence concept summary.You confirm or correct.
2Three sharp questionsAudience, single insight, vision statement. Skipped if the source answers them.Brief assembled.
3Variant + scriptPicks one of three structural variants (customer-led / insight-led / demo-led) and writes a 60–90s script.Script drafted.
3BMotion boardMaps each 5–8s beat to visible action, before/after state, motion mode, and product proof.Motion board approved.
4Design directionChecks for extra visual references, extracts or proposes a visual language, defines the cover-frame strategy, then renders a design thumbnail.Design thumbnail approved.
5Voice auditionFour voices from a curated shortlist of eight, each reading the first 10 seconds.Voice selected.
6Hook renderThe first ~15 seconds are rendered for fast iteration, with frame 0 exported as the cover frame.Hook delivered.
7Iterate → full renderConversational edits. Then the full film.Film complete.

The design thumbnail is the aesthetic iteration unit. The hook render is the video iteration unit. The full render comes after the hook is approved.

Quickstart

Prerequisites

  • Claude with skills support (claude.ai, Claude Code, or any Claude environment that loads skills)
  • Node.js 22+ and FFmpeg for rendering locally
  • Remotion is the default renderer; HyperFrames is available when an HTML/CSS/GSAP project is the better artifact
  • ElevenLabs API key or access to the ElevenLabs Player MCP for voiceover

Install the skill

Option A — .skill file (recommended):

Download studio-video-creator.skill from the latest release and load it into Claude per these instructions.

Option B — clone and link:

git clone https://github.com/clarkvalberg/studio-video-creator.git ~/.claude/skills/studio-video-creator

(See INSTALLATION.md for environment-specific paths.)

Make a film

Open Claude. Attach your source material. Say something like:

"Make a concept film from this brief and prototype. Audience is institutional LPs."

That's it. The skill takes it from there.

What you get

your-project/
├── brief.md           ← interpreted concept, audience, vision
├── script.md          ← final script with section timing
├── motion-board.md    ← beat-by-beat visual causality plan
├── design.md          ← visual / motion language + cover-frame strategy
├── voice.json         ← selected ElevenLabs voice
├── renderer.json      ← selected renderer
├── remotion/ or hyperframes/
│   ├── data/ or src/
│   └── public/
└── out/
    ├── design-thumbnail.png ← Phase 4 · title-frame style artifact
    ├── cover-frame.png ← Phase 6 · actual frame-0 poster image
    ├── hook.mp4       ← ~15s · round-one deliverable
    └── final.mp4      ← 60–90s · 1080p · ship-ready

Everything is editable. Everything is re-renderable. The renderer project is yours — hand it to a designer or engineer to refine, swap in higher-fidelity screens, change the voice for a different audience. The design thumbnail is there so the look can be corrected before voice and video render time enter the loop. The cover frame is there so the actual frozen first impression is judged deliberately, not left to frame-0 luck.

The eight voices

A curated ElevenLabs shortlist keeps voice selection fast. Claude picks four to audition based on your audience and tone, and surfaces the rest only if you ask.

VoiceRegisterFit
RachelWarm intelligent femaleThe default — works across most variants
AdamDeep grounded maleGravitas. LP-facing, institutional
AntoniWarm conversational malePeer-to-peer, developer tools
CharlotteEmotional intimate femaleCustomer-led films, healthcare, care services
BrianConfident American narrator maleCinematic, brand-launch register
LilyWarm British femaleDocumentary cadence, editorial sensibility
DanielBritish authority maleWeighty, regulated industries
WillMeasured thoughtful maleInsight-led films with a contrarian or sophisticated thesis

Full personas and matching matrix in references/voice-shortlist.md.

The banned words

The script rules steer away from these words because they are overused in product and pitch language. They often make a specific concept feel generic before the viewer can hear the idea.

revolutionary · game-changing · best-in-class · cutting-edge · next-generation · leverage · unlock · empower · seamlessly · transformative (self-applied) · innovative (self-applied) · synergy · holistic · robust · scalable (in pitch context) · mission-critical · enterprise-grade · supercharge · AI-powered (use a specific verb instead)

Plus near-banned phrases — "imagine a world where…", "designed to…", "helps you to…" — that the skill steers around. The complete list with reasoning lives in references/script-rules.md.

Format Principles

The workflow draws from short product concept films and studio videos that combine narration, real product surfaces, and a clear visual arc. It is not tied to a specific studio, campaign, or reference.

The useful pattern is simple:

  • Start with a concrete moment, question, or product state
  • Explain the concept through visible change, not only narration
  • Show real or schematic product surfaces
  • Let the voiceover frame the idea without over-selling it
  • Land on a specific promise the concept has earned

Roadmap

The skill stays deliberately narrow: 60–90 second concept films with a gated hook-render loop. The broader studio-video-creator direction is a family of short product videos built from the same workflow:

  • Concept film — current format; source material becomes a voiceover-led pitch film
  • Launch teaser — shorter public-facing reveal for a product or studio project
  • Investor narrative — LP/fundraising cut with sharper market and thesis beats
  • Demo montage — screen-forward product walkthrough with less narration
  • Internal alignment film — strategy narrative for teams before build or launch

New formats should add structure while preserving the current concept-film workflow.

Repository

studio-video-creator/
├── SKILL.md                      ← the operating manual (Claude reads this)
├── references/                   ← deep guidance loaded by phase
│   ├── frameworks.md             ← the canonical structure + 3 variants
│   ├── script-rules.md           ← voice, pacing, banned words
│   ├── motion-board.md           ← beat-by-beat visual causality planning
│   ├── voice-shortlist.md        ← the 8 voices with IDs and fit notes
│   ├── design-language.md        ← UI references, brand extraction
│   ├── cover-frame-strategy.md   ← poster-frame archetypes + checks
│   ├── intake-checklist.md       ← what to extract per input type
│   ├── renderers.md              ← renderer selection guidance
│   ├── hyperframes-integration.md ← HyperFrames layout, data shape, render specs
│   ├── remotion-integration.md    ← Remotion layout and React contract
│   └── example-signatures-law.md ← a complete worked example
├── scripts/                      ← executable helpers
│   ├── init-project.sh
│   ├── audition.sh
│   ├── generate-voiceover.sh
│   ├── check-static-video.sh
│   ├── render-design-thumbnail.sh
│   ├── render-hook.sh
│   └── render-full.sh
├── assets/
│   ├── remotion-template/        ← default React rendering layer
│   ├── hyperframes-template/     ← optional HTML/GSAP rendering layer
│   │   ├── compositions/
│   │   │   ├── design-thumbnail.html
│   │   │   ├── hook.html
│   │   │   └── full.html
│   │   ├── data/
│   │   │   ├── film.json         ← project data contract
│   │   │   └── tokens.json       ← design token contract
│   │   └── scripts/
│   │       └── studio-composition.js
└── docs/                         ← philosophy, workflow, examples

Documentation

Public Reference

This repository is public for reference and installation. External issues, pull requests, discussions, projects, and wiki are disabled. Internal maintenance happens directly in this repository.

License

MIT. See LICENSE.

Built inside Transformative Studio

transformative.studio is the source of truth for current studio context.

Hat tip to Zach Rabin for the spark that nudged this into public.

Made with care. Use it freely.

Ver o original no GitHub

Bom para

Relacionados