GPT Image 2 Skill
leoleoasd/gpt-image-2-skill
Gere e edite imagens com gpt-image-2 via Claude Code, sem sair do terminal
Ver no GitHubSkill para Claude Code que gera e edita imagens usando o modelo gpt-image-2 da OpenAI, direto do terminal. Suporta text-to-image, edição de imagens, inpainting e composição multi-imagem sem dependências externas (usa apenas urllib).
Instalação
npx convez-ai leoleoasd/gpt-image-2-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
- Gerar ilustrações e artes por descrição textual
- Editar e remover elementos de imagens existentes
- Criar logos e ícones minimalistas
- Compor múltiplas imagens em um único resultado
Veredito

Integração prática com gpt-image-2 que reduz fricção para workflows de geração de imagens. Script Python limpo e sem dependências, mas exige créditos de API OpenAI e está acoplado ao ecossistema Claude Code.
README
A tradução para o português deste README ainda está na fila.
gpt-image-2-skill
A Claude Code skill that generates and edits images with OpenAI's gpt-image-2 model — text-to-image, image editing, inpainting, and multi-image composition — without leaving your terminal.
Packaged as a Claude Code plugin so it installs with two commands.
Install
In Claude Code:
/plugin marketplace add luyuxuanleo/gpt-image-2-skill
/plugin install gpt-image-2
Then give it your OpenAI API key. Easiest way — run the setup command and follow the prompt:
/gpt-image-2:setup
It saves the key to ~/.config/gpt-image-2/credentials (permissions 600) so you only do it
once. Prefer to manage it yourself? Either of these also works:
export OPENAI_API_KEY=sk-... # per shell session
printf '%s' sk-... | python3 plugins/gpt-image-2/skills/gpt-image-2/gpt_image.py set-key -
The skill checks OPENAI_API_KEY first, then the saved file.
Use
Just ask Claude, e.g.:
- "Generate a watercolor red fox sleeping in the snow, portrait orientation."
- "Make a transparent PNG logo of a minimalist mountain icon."
- "Edit
photo.pngto look like a vintage oil painting." - "Combine
product.pngandbeach.pnginto one product shot on the beach."
Claude invokes the skill, saves the image next to your working directory, and gives you the path.
Run the script directly
The skill is a single zero-dependency Python script (urllib only — no pip install):
python3 plugins/gpt-image-2/skills/gpt-image-2/gpt_image.py \
generate "a red fox in snow" --size 1024x1536 --quality high --out fox.png
python3 plugins/gpt-image-2/skills/gpt-image-2/gpt_image.py \
edit "make the sky a starry night" --image scene.png --mask sky.png --out night.png
Options
| Option | Values |
|---|---|
--size | WxH or auto — edges multiple of 16, max 3840, ratio ≤ 3:1 (e.g. 1024x1024, 2048x2048, 3840x2160) |
--quality | low, medium, high, auto |
--background | transparent, opaque, auto |
--format | png, jpeg, webp |
--compression | 0-100 (jpeg/webp) |
--n | number of images |
--out | output path |
--model | override model id (or set GPT_IMAGE_MODEL) |
Configuration
| Env var | Purpose |
|---|---|
OPENAI_API_KEY | required — your OpenAI API key |
GPT_IMAGE_MODEL | optional — override the model (default gpt-image-2) |
OPENAI_BASE_URL | optional — point at a compatible/proxy endpoint |
License
MIT