OpenAI has just officially introduced a remarkable upgrade to the AI image generation capability in ChatGPT, an important step instead of using a separate image generation model like the previous DALL-E. This new feature has been integrated directly into GPT-4o, bringing about remarkable improvements.
Overcoming inherent limitations
While many current AI image generation models can create impressive artistic images, they often struggle with elements like text, logos, and everyday objects. OpenAI claims that its new GPT-4o can overcome these limitations by:
- Display text correctly
- Strictly adhere to user requirements
- Leverage background knowledge and conversational context
- Allows editing of uploaded photos or creation of new photos based on original photos
- Widely available
This new feature is currently rolling out to ChatGPT Free, ChatGPT Plus, Pro, and Team users, and will be available to ChatGPT Enterprise and Edu in the coming weeks. Notably, this will be the default image creation tool in ChatGPT, making it easy for users to access without any additional options. Users can customize images with:
- Specific aspect ratio
- Exact color (using hex code)
- Transparent background
- Multi-platform support

In addition to ChatGPT, this feature will also be available on platforms including Sora (image generation), dedicated DALL·E GPT, and GPT-4o API (for developers, launching in the coming weeks).
Despite its promise of many improvements, the new model still has some limitations:
- Image generation time can be up to 1 minute due to high detail
- Unwanted cropping with vertical photos
- Sometimes "fabricate" information with little context required
- Difficulty processing more than 10-20 concepts at once
- Difficulty with non-Latin languages
- Detailed corrections (like spelling errors) are not very effective
- Difficult to display detailed information at small sizes
All images generated by GPT-4o will contain C2PA metadata, allowing provenance verification using OpenAI's internal tools.
Despite some limitations, GPT-4o promises to produce more accurate and customized images. OpenAI says it will continue to improve the model in the coming months, opening up new possibilities for AI-powered visual content creation.
With this major update, OpenAI continues to strengthen its leadership in the creative AI race, delivering a more seamless and powerful experience to users across multiple platforms.