OpenAI has launched a new image generation feature integrated directly into its GPT-4o model, allowing users to create more accurate and contextually relevant visuals through natural conversation.
In an official announcement, OpenAI said, “GPT‑4o image generation excels at accurately rendering text, precisely following prompts, and leveraging 4o’s inherent knowledge base and chat context—including transforming uploaded images or using them as visual inspiration.” The company called the upgrade a step toward making AI visuals a practical tool with "precision and power."
Photo: Testing out the CPT-4o
Key features and capabilities
The newly integrated image generation system enables users to:
- Accurately render text within images
- Maintain consistent visual styles through iterative conversation
- Handle complex prompts with up to 20 distinct elements
- Generate visuals inspired by uploaded references
- Create images using GPT-4o’s built-in training data and conversation history
A major advantage of the update is its conversational refinement ability. For instance, if a user is designing a video game character, GPT-4o can maintain visual consistency across different iterations of the same character while applying incremental changes based on user feedback.
Known limitations Despite the powerful capabilities
OpenAI has acknowledged some current limitations in the system:
- Cropping issues: Tall images like posters may be clipped at the bottom.
- Prompt hallucinations: Vague prompts can result in inaccurate or misleading visuals.
- Blending errors: The model struggles with overly dense prompts (e.g., a full periodic table).
- Multilingual text challenges: Non-Latin scripts may not render correctly.
- Editing constraints: Isolated edits to specific image parts may unintentionally alter other areas.
- Facial consistency issues: Uploaded images with faces may not maintain likeness across edits.
- Information density problems: Small visuals may lose important detail.
OpenAI says these are known issues and will be addressed through future model improvements.
Implications for web and search
With this upgrade, AI-generated visuals move from decorative novelty to practical tool—especially for business, design, and communication.
OpenAI noted that all images include C2PA metadata for transparency. The company also encourages best practices such as providing alt text, using images to support user intent, and avoiding generic, template-style designs.
Photo: Testing out the GPT-4o
While Google does not penalize AI-generated images in search rankings, Search Advocate John Mueller has previously voiced skepticism over their usefulness. Google is also working to label AI-generated content in its search results for transparency.
Access and availability
The image generation feature is now live for ChatGPT users on the Free, Plus, Pro, and Team plans. Enterprise and Education accounts will get access soon. Developers can expect API access in the coming weeks.
Due to higher processing needs, each image may take about a minute to generate.
COMMENTS
Comments are moderated and generally will be posted if they are on-topic and not abusive.
For more information, please see our Comments FAQ