OpenAI has released a major update to its Artificial Intelligence image generator, ChatGPT Images. The 2.0 upgrade, introduced in a blog post on April 21, adds “thinking capabilities” for the first time. The new function lets the imaging model search the internet for real-time information using a single prompt, then create multiple images and double-check its own outputs.
OpenAI said the ability to think allows the system to do “more of the heavy lifting” between idea and image, with stronger accuracy and visual cohesion. The update also reflects more current information because of a knowledge cut-off of December last year, when OpenAI rolled out its last big Images update. ChatGPT’s handling of more sophisticated imagery is also supported by better rendering of small text and iconography, along with improved performance on dense compositions.
In thinking mode, users can create up to eight images at once, a first for ChatGPT. OpenAI positioned this as useful for more complex creative work such as social media graphics in different aspect ratios and languages, or a family of poster concepts. The model also puts more emphasis on languages beyond English and those using Latin script. It now supports Japanese, Korean, Chinese, Hindi and Bengali.
OpenAI said photo outputs now capture the “tiny flaws that add realism,” and the tool has become more capable of depicting a wider range of styles. The company highlighted cinematic stills, manga and pixel art as examples, aimed at uses including marketing and gaming. A wide array of aspect ratios is available, ranging from 3:1 to 1:3.
The upgraded Images is now available to all ChatGPT users. Coders can access it through the Codex app, while developers and businesses can use the gpt-image-2 model in the API, where pricing depends on image quality and resolution. Advanced outputs with thinking are available to Plus, Pro and Business users. OpenAI also said that in the API, outputs over 2K are in beta and may produce inconsistent results.