OpenAI updates ChatGPT Images 2.0 to use web info and make image sets

In short: OpenAI has rolled out ChatGPT Images 2.0, an updated image maker that can use web information and create multiple related images from one request.

What happened

OpenAI says it is releasing the latest version of its image generator inside ChatGPT, called ChatGPT Images 2.0. The company says the tool is better at following instructions, keeping specific details you ask for, and placing readable text inside images.

A key change is a new optional “thinking” mode. When this is turned on, the image tool can pull information from the web to help build the image. Think of it like letting the tool quickly look things up before it starts drawing.

With “thinking” enabled, ChatGPT Images 2.0 can also make up to eight images at once from a single prompt, while keeping the same characters, objects, and style across the set. OpenAI says this can help with things like manga pages, a group of social media graphics, or a room by room design plan for a house.

OpenAI also says all users benefit from other upgrades, including better handling of photo likeness, more styles like pixel art and cinematic stills, and higher image quality up to 2K (similar to a sharper, higher resolution screen). The tool also supports more image shapes, from very wide to very tall, and it is improved at rendering text in languages including Japanese, Korean, Chinese, Hindi, and Bengali.

Why it matters

For regular users, this could mean less back and forth to get a consistent set of images, like a simple storyboard or matching marketing posts. It also raises a practical question to watch, which is how clearly the tool shows what web sources it relied on when it used online information.

Source: The Verge AI