Is GPT Image 2 free with this skill?

There is no extra cost beyond your ChatGPT subscription, but generations count toward your plan's usage limits.

Image generation for Claude Code, powered by GPT Image 2

Q: Can Claude generate images?

Not natively. Claude can read images but not create them. This free Claude Code skill adds image generation by calling GPT Image 2 through the Codex CLI.

Q: Do I need an OpenAI API key for GPT Image 2?

No. The skill uses the Codex CLI signed in with your ChatGPT account, so images count against your ChatGPT plan instead of pay-as-you-go API billing.

Q: How do I install skills in Claude Code globally?

A skill is a folder with a SKILL.md file. Put it in ~/.claude/skills to use it in every project, or in .claude/skills inside one project. The one-line installer does the global install for you.

Q: Does it work in claude.ai or the Claude Desktop chat?

It is built for Claude Code (terminal, desktop app and IDE extensions), which can run local commands. The web chat cannot run the Codex CLI.

A free skill that lets Claude create and edit images on your ChatGPT subscription. No API key. One command.

Install free Star on GitHub

Vintage Vancouver travel poster with a red seaplane, generated with GPT Image 2 in Claude Code

Product photo of a sage-green ceramic mug generated by Claude Code with GPT Image 2

Watercolor of a plant nursery stall at a spring market, generated with GPT Image 2

Install the skill in Claude Code

One command installs it globally, for every project. The installer can also set up the Codex CLI and sign you in with ChatGPT.

$ curl -fsSL https://raw.githubusercontent.com/mishagavura/claude-code-image-generation/main/install.sh | bash

$ npx -y github:mishagavura/claude-code-image-generation

› /plugin marketplace add mishagavura/claude-code-image-generation
› /plugin install codex-image@claude-code-image-generation

› Install the skill from https://github.com/mishagavura/claude-code-image-generation globally

Global by default: ~/.claude/skills One project only: add | bash -s -- --project Restart Claude Code when it finishes

What Claude can make with GPT Image 2

Every image on this page was generated by the skill from the prompt shown under it. No retouching.

Rainy Tokyo street at night with neon reflections, generated by GPT Image 2 — Rainy Tokyo side street at night, neon reflecting on wet asphalt, 35mm film look

Overhead food photo of a sourdough loaf generated with GPT Image 2 — Overhead photo of a rustic sourdough loaf and butter on linen, morning light

Isometric illustration of a cozy developer desk with a cat, made in Claude Code — Isometric 3D illustration of a cozy developer desk with a sleeping cat

Website hero for a coffee brand, headline "Slow coffee, fast mornings"

Retro Vancouver travel poster with a seaplane generated by GPT Image 2 — 1950s travel poster, red seaplane over the harbour, title "VANCOUVER"

Studio product shot of a green ceramic mug made with GPT Image 2 — Studio product photo of a matte sage-green mug on pale oak, window light

Can Claude generate images? Now it can.

Claude can read images but can't create them. This skill hands the drawing to GPT Image 2 through the Codex CLI, and Claude stays in charge of the result.

You ask in plain English
"Make a hero image for the homepage." Or type /codex-image.
Claude writes the prompt
Subject, style, framing and exact text, based on your project.
GPT Image 2 renders it
Through Codex CLI on your ChatGPT plan, in about 30 to 90 seconds.
Claude checks it and saves it
It looks at the image, retries if something is off, and puts the file where your code needs it.

› make a landscape hero image for the coffee landing page

⏺ Skill(codex-image)
⏺ Bash gen.sh -o public/hero.png -a landscape \
    "Website hero for a specialty coffee brand,
     headline 'Slow coffee, fast mornings' ..."
  ✓ /app/public/hero.png

⏺ Read public/hero.png
  Headline spelled correctly, bag and beans on
  the right, room for the nav. Looks good.

⏺ Saved to public/hero.png and added it to
  Hero.tsx.

More than text-to-image

Edit the images you already have

Change color, lighting or season, remove objects or swap backgrounds. Only what you ask for changes.

Original oak leaf illustration before editing — Before

Oak leaf edited to autumn red with GPT Image 2 — "Make it autumn red"

No OpenAI API key

Runs on the ChatGPT plan you already pay for.

Up to 4 variants

Generated in parallel, so you can pick the best one.

Reference images

Match the style or composition of an image you have.

Transparent PNGs

Stickers, cutouts and icons with a real alpha channel.

Generated image saved into a web project

Saved into your project

Claude puts the file in place and references it in your code.

Saved you an API bill? Leave a star.

It's free and open source. A star takes one click and helps other Claude Code users find it.

Star on GitHub

FAQ

Can Claude generate images?

Not natively. Claude can read and analyse images but can't create them. With this skill, Claude Code generates and edits images by calling GPT Image 2 (gpt-image-2) through the Codex CLI.

Do I need an OpenAI API key for GPT Image 2?

No. The Codex CLI signs in with your ChatGPT account, so images count against your ChatGPT plan instead of pay-as-you-go API billing.

Is GPT Image 2 free with this?

There's no extra cost beyond your ChatGPT subscription, but each image counts toward your plan's usage limits. When you hit the limit, the skill tells you instead of failing silently.

How do I install skills in Claude Code globally?

A skill is a folder with a SKILL.md file. Put it in ~/.claude/skills/ to use it in every project, or in .claude/skills/ inside one project. The installer above does the global install for you.

Does it work in claude.ai or Claude Desktop chat?

It's built for Claude Code (terminal, desktop app and IDE extensions), which can run local commands. The web chat can't run the Codex CLI.

Which operating systems are supported?

macOS and Linux. On Windows, use WSL. The Codex CLI needs Node.js 18 or newer.