OpenAI readies Custom GPTs for native four-dimensional image creation
Update on ChatGPT’s Image Generation Capabilities
A recent enhancement to the ChatGPT web interface has introduced a new feature called "GPT‑4o Image Generation" within the Custom GPT builder. This advancement indicates that OpenAI is gearing up to allow user-created bots to generate images using the same advanced multimodal model that powers ChatGPT’s image functionality.
Overview of the New Feature
The new image generation option was first identified during an update on April 18, 2025. This setting replaces the previous DALL‑E 3 backend, which suggests that when the feature is officially launched, it will provide similar capabilities to the existing image tools.
If activated, Custom GPTs will be able to directly transform user prompts into photorealistic or stylistically unique visuals without needing to use an external image service. This change aims to streamline creative workflows, enabling specialized assistants like fashion stylists or biology tutors to generate, modify, and refine images within a single conversation. This integration makes for a more interactive and engaging experience for users.
Enhanced Performance with GPT‑4o
The image generation model associated with GPT‑4o, which was rolled out to the main ChatGPT product on March 25, 2025, is known for delivering sharper text rendering and improved character portrayal. Unlike previous models, it uses an autoregressive method for image creation instead of the traditional diffusion approach. This innovative technique enhances accuracy, although it may require a little more time to process the images.
Key Features of the GPT‑4o Image Generation
- Precision Image Generation: The autoregressive technique offers significant improvements in the quality of images produced.
- Consistent Character Depiction: The model ensures that characters are represented consistently across images.
- Stronger Attribute Binding: This feature allows for better integration of visual elements, contributing to overall coherence.
Experimenting with Custom GPTs
Developers and enterprises can leverage the new image generation capabilities to enhance their Custom GPTs, leading to exciting opportunities for user engagement. One example is the "Cyberpunk Artist" GPT, which has enabled 4o image generation. While this feature is still in the experimental phase, it showcases the potential of the new technology.
Users should be aware that the experimental model has shown a success rate of approximately 50% in following prompts correctly during testing.
Try It Out!
For those interested in exploring this new feature, you can access the "Cyberpunk Artist" GPT and experience the image generation in action. Navigate to the link below to see how it performs:
Implications for Developers and Enterprises
Bringing the advanced capabilities of GPT‑4o image generation to Custom GPTs could significantly benefit developers and businesses. Additionally, this upcoming support hints at the possibility of a unified multimodal stack across ChatGPT, Sora, and the API, which would ensure a consistent experience and reduce fragmentation in the available tools.
While OpenAI has not yet specified a release timeline for the full rollout of this feature, the presence of the in-product toggle indicates that a gradual implementation is likely to follow. Once everything is operational, developers should find it easy to switch existing bots to utilize GPT‑4o image generation, leading to enriched user interactions and a more dynamic overall experience.
