A new option — “GPT‑4o Image Generation” — has appeared in the web interface of ChatGPT for creating images in the custom GPT builder. This indicates that OpenAI is preparing to launch the ability for user-created bots to generate images using the same multimodal model already employed in ChatGPT’s own image tool. The new feature replaces the previous DALL‑E 3 support, which may provide equal capabilities for all users once the rollout begins.
Thanks to this change, custom GPTs will be able to convert text prompts directly into photorealistic or stylized visual images without using third-party services. This will allow various specialized assistants — such as stylists or biology teachers — to create, edit, and refine images within a single conversation, making interactions with bots more dynamic and convenient for users.
The GPT‑4o model for working with images, introduced in ChatGPT at the end of March, demonstrates improved text rendering, more accurate depiction of object features, and ensures consistency in character images. Images are generated step by step, which increases accuracy, though it takes a bit more time compared to previous approaches.
The new feature can already be tested with the “Cyberpunk Artist” GPT, which has image generation via GPT‑4o enabled. For now, this function is at an experimental stage and results may be inconsistent, but expanding such capabilities for custom GPTs opens up additional advantages for developers and businesses.
OpenAI has not yet announced a specific launch date, but the appearance of the corresponding toggle usually precedes gradual rollout for Plus, Pro, and Team users. Once deployment begins, bot creators will be able to quickly enable GPT‑4o image generation, increasing user engagement and expanding their assistants’ functionality.
