OpenAI’s GPT-4o can now create better AI images

OpenAI has just rolled out a major update to its AI model, GPT-4o, bringing image generation directly into ChatGPT. This means users can now create detailed and lifelike images simply by describing what they want. The feature is already available for free and paid users of ChatGPT, making it easier than ever to turn ideas into visuals.

What’s New?

GPT-4o isn’t just another AI image generator — it’s built into ChatGPT, allowing seamless text-to-image creation. Whether you need artwork, posters, memes or infographics, you can generate them instantly. A key improvement is better text rendering within images. Unlike older AI models that struggled to place readable text inside images, GPT-4 apparently does this much more accurately now.

In the latest version, users can refine images through natural conversation. For example, if you’re designing a video game character, the character’s appearance remains coherent across multiple iterations as you refine and experiment. It can also now analyse and learn from user-uploaded images, seamlessly integrating their details into its context to inform image generation.

How safe is it?

OpenAI says that they have built-in safety measures to prevent misuse. The system blocks inappropriate content and doesn’t generate real people’s images without permission. Plus, every AI-created image has a digital watermark, so you can always tell if an image was AI-generated.

Who can use it?

The feature is currently rolling out to all ChatGPT Free, Plus, Team, and Pro users. Free users will have some limitations on how often they can generate images, while paid users get more access and higher-quality outputs.

While the tool is impressive, it’s not perfect yet. In some cases, it might miss small details in complex prompts or struggle with very specific requests.

A couple of weeks ago, OpenAI also launched new speech-to-text and text-to-speech audio models in its API, making it possible to build more customisable, and intelligent voice agents. The aim was to improve the accuracy of voice agents in challenging scenarios involving worldwide accents, noisy environments and varying speech speeds. These improvements increase transcription reliability, making the models especially well-suited for use cases such as customer call centres, meeting transcriptions, and more.