imagegenlisted

# imagegen Generate and edit images for the current project with **`gpt-image-2` only**, exclusively through the **Codex CLI** (`codex exec`) and its built-in `image_gen` tool, billed on the user's **ChatGPT subscription**. ## Requirements - Codex CLI installed (`codex --version`), on a recent version whose agent exposes the built-in `image_gen` tool. - Logged in with a ChatGPT account: `codex login`. - For transparent cutouts only: Python 3 with Pillow (`python3 -m pip install pillow`). ## When to use - Generate a new image: concept art, product shot, hero/banner, game asset, infographic. - Generate using one or more reference images for style, composition, or mood. - Edit an existing image: object removal/replacement, lighting or weather changes, background replacement, compositing, text localization, sketch-to-render. - Produce a transparent-background cutout (via the chroma-key pipeline in [references/transparency.md](./references/transparency.md)). - Produce several variants of one asset. ## When not to use - Extending or matching an existing SVG/vector icon set, logo system, or illustration library in the repo — edit those natively. - Simple shapes, diagrams, wireframes, or icons better produced in SVG, HTML/CSS, or canvas. - Any task where the user wants deterministic code-native output rather than a generated bitmap. ## gpt-image-2 model notes - Strong instruction following, layout control, and in-image text rendering — quote exact text verbatim in the prompt