OpenAI has announced the integration of image generation directly into ChatGPT, allowing users to create images within the chatbot using GPT-4o.
Previously, users relied on DALL·E to generate images, either through a separate platform or within ChatGPT. Now, OpenAI has integrated an even more advanced image generation capability directly into ChatGPT using GPT-4o.
The feature was rolled out on Tuesday, March 25, across Plus, Pro, Team, and Free subscription tiers.
The company disclosed this in a press release, stating that the feature represents a significant step in making image creation an essential part of AI-driven communication.
Sam Altman, OpenAI’s CEO, on his X page, described it as “an incredible technology/product,” emphasizing that the feature marks a new level of creative freedom, acknowledging that while people will create remarkable content, some outputs may offend people.
“Two things to say about it: 1. It’s an incredible technology/product. I remember seeing some of the first images come out of this model and having a hard time, they were really made by AI. We think people will love it, and we are excited to see the resulting creativity. Secondly, this represents a new high-water mark for us in allowing creative freedom. People are going to create some really amazing stuff and some stuff that may offend people; what we’d like to aim for is that the tool doesn’t create offensive stuff unless you want it to, in which case within reason it does. as we talk about in our model spec, we think putting this intellectual freedom and control in the hands of users is the right thing to do, but we will observe how it goes and listen to society. We think respecting the very wide bounds society will eventually choose to set for AI is the right thing to do, and increasingly important as we get closer to AGI. Thanks in advance for the understanding as we work through this,” he stated.
New capabilities in GPT-4o’s image generation
Unlike previous iterations, GPT-4o’s image generation is designed to be more precise, flexible, and interactive. OpenAI stated that the model excels in several key areas:
- Text Rendering: Unlike past AI models that struggled to place readable text within images, GPT-4o generates clear and accurate text, making it suitable for diagrams, infographics, and labeled visuals.
- Multi-turn Generation: Users can refine images through conversation, allowing step-by-step modifications while maintaining consistency across versions. This feature benefits tasks like character design, storyboarding, and branding.
- Instruction Following: GPT-4o can handle detailed prompts with greater accuracy, generating images with up to 10-20 distinct objects while maintaining their relationships and traits.
- In-Context Learning: The AI can analyze uploaded images and incorporate details into new creations, making it useful for design inspiration and visual brainstorming.
- Knowledge Integration: The model links its understanding of text and images, allowing it to generate context-aware visuals, such as weather infographics, technical diagrams, and educational illustrations.
More insights
OpenAI also revealed that developers will gain API access in the coming weeks, enabling broader integration of GPT-4o’s image capabilities across applications.
- The image generation feature is rolling out immediately for ChatGPT users, with planned expansion to Enterprise and Education subscribers.
- Users can generate images by simply describing their requirements in ChatGPT, with options to specify colors, aspect ratios, and other design elements.
Due to the complexity of the model, OpenAI noted that rendering images may take up to a minute.
For developers and businesses looking for more customization, OpenAI confirmed that access to DALL·E will still be available as a separate model option.