convez.ai
SkillDesign & UIIntermediário
guyaga

Comics Maker

guyaga/10d10s-bonus-comics-maker

Crie quadrinhos print-ready com arte IA consistente em minutos

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill que transforma um conceito e uma foto sua em uma história em quadrinhos completa, pronta para impressão. Planeja, gera arte com Gemini 3 Pro (mantendo consistência facial em cada painel), adiciona textos e vincula tudo em PDF. Ideal para criar narrativas visuais coerentes onde o personagem principal permanece reconhecível do início ao fim.

Instalação

$npx convez-ai guyaga/10d10s-bonus-comics-maker

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
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Gerar histórias em quadrinhos personalizadas com seu rosto como protagonista
  • Criar portfólios visuais ou narrativas de marketing em formato de HQ
  • Produzir material de treinamento ou storytelling empresarial em quadrinhos

Veredito

Ferramenta bem arquitetada que resolve o problema real de consistência facial em gerações repetidas de IA. Pipeline claro (planejar → gerar → letreirar → vincular) e integração nativa com Claude Code a tornam acessível. Útil para criadores de conteúdo e aplicadores em narrativa visual.

README

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

📖 Bonus: Comics Maker

10 Days 10 Skills — course bonus Turn a one-line concept (and your own face) into a full, print-ready comic book. Plan → generate art → letter → bind PDF. The hero stays recognizable on every page.

This is the skill behind "The Debugger" — a 12-page cyberpunk issue where a real person becomes a code-bending hero in Tel Aviv 2049. You describe the comic in one config file; three small Node scripts do the rest. A reference photo is fed into every generation call, so the face, hair, build and signature outfit stay consistent across all twelve pages.

It builds on the course's image day:

StepToolCourse day
Plan the issueCOMIC_PLAN.md (character sheet + page-by-page script)—
Generate the artGemini 3 Pro Image (fast) or GPT Image via the Codex CLI (sharper, free on a ChatGPT login) — photo on every panelDay 1: Image Generation
Letter the pagessharp + hand-built SVG (classic white caption boxes, speech bubbles, SFX)—
Bind the bookpdfkit → US-comic-sized PDF—

Two art backends (backend: "gemini" | "gpt2")

  • gemini (default) — Gemini 3 Pro Image. Fast (~seconds/page). Needs GEMINI_API_KEY.
  • gpt2 — drives the Codex CLI's built-in image_gen tool (codex exec -m gpt-5.5). Noticeably sharper line work and crisper in-art text (titles, SFX, signage), ~1–2 min/page, and free on a ChatGPT-account codex login (no OPENAI_API_KEY billed). The reference photo is passed with -i so the hero's face carries across every page.

🔁 The pipeline

plan (COMIC_PLAN.md) → comic.config.mjs → generate.mjs → assemble.mjs → make-pdf.mjs → <title>.pdf
  1. Plan — interview the user, write COMIC_PLAN.md (concept, hero, art style, page breakdown).
  2. Config — author comic.config.mjs: a stylePrefix + charDesc reused on every prompt, then one entry per page (prompt + text overlays).
  3. Generate — scripts/generate.mjs makes the character sheet + every page, feeding the reference photo to each call. Re-run a single page with --only=page04.
  4. Letter — scripts/assemble.mjs composites caption boxes / speech bubbles / SFX onto the art (pages where the model already drew the text just get a page number).
  5. Bind — scripts/make-pdf.mjs lays every page onto a comic-sized PDF.

🚀 Quick start

git clone https://github.com/guyaga/10d10s-bonus-comics-maker
cd 10d10s-bonus-comics-maker
cp .env.example .env          # add your GEMINI_API_KEY
npm install

# set up a comic project
mkdir -p "My Comic/Assets"
cp reference/comic.config.mjs "My Comic/comic.config.mjs"   # then edit it
# drop your hero photo at "My Comic/Assets/hero-reference.jpg"

export GEMINI_API_KEY=your-key      # PowerShell: $env:GEMINI_API_KEY="your-key"

# run the three stages
node scripts/generate.mjs "My Comic"
node scripts/assemble.mjs "My Comic"
node scripts/make-pdf.mjs "My Comic"
# -> "My Comic/The_Debugger_Issue_1.pdf"

Installed as a Claude Code skill (in ~/.claude/skills/comics-maker/), just say "make me a comic about …" and Claude runs the whole pipeline for you.


🔑 Keys

Key / loginForRequired?
GEMINI_API_KEYthe gemini backend (Gemini 3 Pro Image)for gemini
codex CLI logged in (ChatGPT account)the gpt2 backend (free built-in image_gen)for gpt2
OPENAI_API_KEYonly gpt2 with mode:"api" (paid gpt-image-2)optional

🧱 Config in one screen

comic.config.mjs exports the whole comic. Define the two consistency constants once, reuse them everywhere:

const stylePrefix = "Modern graphic novel comic book art style. ...";
const charDesc    = 'The main character looks exactly like the reference photo: ...';

export default {
  title: "The Debugger - Issue 1",
  refPhotos: ["Assets/hero-reference.jpg"],   // fed into every page
  defaultAspect: "3:4",
  pageSize: { widthIn: 6.625, heightIn: 10.25 },
  characterSheet: { file: "sheet.png", aspect: "16:9", prompt: `${stylePrefix} ...` },
  pages: [
    {
      id: "page01", folder: "Page_01", file: "page01.png",
      useRef: false,                          // no hero on this page
      prompt: `${stylePrefix} FULL PAGE SPLASH ...`,
      overlays: [{ type: "caption", text: '"Tel Aviv. 2049."', x: 0.03, y: 0.82, width: 820 }],
    },
    // ...one entry per page
  ],
};

Overlay types: caption (purple narrator box), speech (white bubble w/ tail), sfx (red glitch text), title, subtitle. Positions are fractions of the page (0–1).

See reference/comic.config.mjs for the complete 12-page "Debugger" example and reference/COMIC_PLAN.md for the planning doc that precedes it.


💡 Hard-won lessons (baked into the skill)

  • Feed the reference photo to every hero panel. This is the single biggest lever on face consistency — far more than describing the face in words.
  • Reuse the style prefix verbatim on every prompt, or the art style drifts page to page.
  • Let the model draw the big lettering (SFX, title cards, signage) inside the art, and add only clean narrator captions / speech bubbles afterward. Mixing both reads best.
  • Position overlays with fractions, not pixels, so they survive any output resolution.
  • Splash pages (1, 7, 12) carry the issue — spend the most prompt care there.
  • A page came out wrong? Tweak its prompt and re-run just that page with --only= — no need to regenerate the whole issue.

Course bonus for 10 Days, 10 Skills. MIT licensed.

Ver o original no GitHub

Bom para

Relacionados