OpenAI’s New Image Generation Feature Allows You to Create Images Directly in ChatGPT

OpenAI has introduced a new image generation feature in GPT-4o, allowing ChatGPT users to create images directly from the chatbot. OpenAI CEO Sam Altman announced this development on social media, describing it as an "incredible technology/product" that offers users enhanced creative control. Altman shared his initial disbelief at the AI-generated images' authenticity, highlighting the tool's potential for broad creative expression.

In a blog post, OpenAI detailed that GPT-4o excels in rendering text accurately and maintaining consistency across multiple iterations. The model is designed to support various applications, including game development and educational materials. It can handle up to 10-20 different objects in a single image, showcasing significant improvements over previous models.

Image Generation Tool Features

The new tool allows users to refine images through natural conversation, enabling seamless adjustments to elements like character designs. Unlike earlier models that struggled with precise details and object relationships, GPT-4o generates images with higher accuracy and contextual awareness. This advancement marks a significant leap in AI image generation capabilities.

OpenAI has implemented improved safety features to ensure responsible use of the image generator. All AI-generated images will include metadata via C2PA, indicating their creation using GPT-4o. Additionally, an internal search tool has been developed to verify content and detect AI-generated visuals, preventing harmful or policy-violating images such as deepfakes.

Limitations and Future Improvements

Despite its advancements, the model still faces certain limitations. It struggles with rendering non-Latin languages accurately and may crop images incorrectly, especially longer ones like posters. Additionally, it might generate inaccurate details when dealing with complex images or editing specific portions without unintended changes.

To address these challenges, OpenAI is working on improvements such as better precision for editing images and enhancing consistency in facial features when modifying user-uploaded images. The company is also focusing on improving how the model processes small details and handles intricate compositions.

Availability and Future Access

The image generation feature is currently being rolled out to Plus, Pro, Team, and Free users in ChatGPT. Enterprise and Education users are expected to receive access soon. OpenAI also confirmed that developers will be able to use the feature via the API in the coming weeks.

"We think putting this intellectual freedom and control in the hands of users is the right thing to do," Altman stated. However, he emphasised that OpenAI will closely monitor usage and adjust policies as needed while respecting societal boundaries as AI progresses towards AGI.

Source