Using AI to generate images isn’t new. However, in the past, AI-generated images had a lot of telltale signs. They were also notoriously unreliable, where sometimes the changes you requested were ignored or misunderstood. That has changed with the new ChatGPT Images tool, which based on the demonstrations, looks pretty amazing .
New ChatGPT Images
OpenAI has announced a new version of ChatGPT Images, and it’s quite a substantial update. According to the company, the new version is up to 4x faster, and it has the ability to make precise edits.
The company says, “Now, when you ask for edits to an uploaded image, the model adheres to your intent more reliably—down to the small details—changing only what you ask for while keeping elements like lighting, composition, and people’s appearance consistent across inputs, outputs, and subsequent edits.”
Users can request the AI model to add, remove, combine, and blend multiple images together. It can even transform images by using multiple photos and turning it into something new entirely. OpenAI claims that the new model can follow instructions more reliably and can even render text precisely.
Also, if you’re generating images where there might be a lot of small faces in the background, like a crowded street or a stadium, the new model seems to be able to do a much better job of that. If you want to check out the new features, it’s rolling out now to all ChatGPT and API users globally. OpenAI reassures users that there are still a lot of improvements to be made, but for now, the new version seems to be a lot better.
Going up against Nano Banana
So far, the demonstrations and examples OpenAI gave are impressive. However, OpenAI has some pretty stiff competition from Google’s Nano Banana . There are multiple threads on Reddit where users compare the differences between the images generated by OpenAI and Google, and honestly, they’re both pretty good.
