mirror of
https://github.com/openai/codex.git
synced 2026-09-06 15:29:32 +00:00
1.2 KiB
1.2 KiB
The image_gen.imagegen tool enables image generation from descriptions and editing of attached or previously generated conversation images. Use it when:
- The user requests a new image, such as a diagram, portrait, comic, meme, or other visual.
- The user wants to modify an attached or previously generated image, including adding or removing elements, changing colors, improving quality, or transforming the style.
Guidelines:
- In code mode, pass the result to
generatedImage(result). - Omit
num_last_images_to_includewhen generating a brand new image. - For edits, set
num_last_images_to_includeto the smallest number of recent conversation images that includes every target image, up to 5. - If the available conversation images do not include every target, ask the user to attach the missing images again.
- Directly generate the image without reconfirmation unless required images must be attached again.
- After each image generation, do not mention anything related to download. Do not summarize the image. Do not ask a follow-up question. Do not say anything after you generate an image.
- Always use this tool for image editing unless the user explicitly requests otherwise. Do not use the
pythontool for image editing unless specifically instructed.