OpenAI releases 'ChatGPT Images' with the image generation model 'GPT Image 1.5,' which generates images up to four times faster than Google's Nano Banana Pro
On December 17, 2025, OpenAI released a new version of ChatGPT Images , equipped with a new image generation model called GPT Image 1.5 . This new model is designed to more reliably reflect user intent not only when generating images from scratch, but also when editing existing photos.
The new ChatGPT Images is here | OpenAI
OpenAI's new flagship image generator AI is here | The Verge
https://www.theverge.com/ai-artificial-intelligence/845558/openais-new-flagship-image-generation-model-gpt-image-1-5
GPT Image 1.5 allows for precise editing, adding, removing, combining and blending elements while maintaining consistency in lighting, composition and people's appearance, allowing for changes without compromising the original image. It also generates images up to four times faster.
Below is the result of loading a photo of OpenAI CEO Sam Altman and entering the prompt 'Transform me into an iconic 1980s VHS fitness instructor,' along with detailed instructions. The image quality reminiscent of 1980s VHS is faithfully reproduced, and the caption 'LET'S GET PHYSICAL' is intact.
Below is an image mixed with Vermeer's 'Girl with a Pearl Earring.'
This is what it looks like when made into an enamel Christmas ornament. It's faithful to the prompt, and the lighting and composition are natural, with no major flaws.
Below are the results of giving instructions to GPT Image 1.5 (left) and the previous model (right) to 'Draw a 6x6 grid and draw an illustration as instructed within that grid.' From the instruction to draw the initial 6x6 grid, it can be seen that GPT Image 1.5 is more accurate. OpenAI states that 'the relationships between elements are maintained as intended, allowing for more accurate editing and more complex original compositions.'
In addition, text rendering capabilities have been enhanced, allowing for more accurate rendering of denser, smaller text and Markdown. Below is a newspaper-style image generated by GPT Image 1.5 that summarizes the features and benchmark results of GPT-5.2. While traditional image generation AI often renders characters in images with distorted shapes, rendering them non-characteristic, the images generated by GPT Image 1.5 accurately display small characters.
In addition, the depiction of many small faces and overall naturalness have also been improved. While clear improvements can be seen in examples such as the deep-sea creature poster, OpenAI says that it is still not perfect in some areas, such as the generation of specific styles and multiple faces, and there is room for further improvement.
For businesses and developers, the API is available as 'GPT Image 1.5,' with input and output costs 20% lower than the previous model, GPT Image 1. This model excels at maintaining consistency in brand logos and products, making it suitable for creating marketing materials and product catalogs, and is already being used by companies such as Wix and Canva.
ChatGPT Images is currently rolling out to all ChatGPT users and API users, with business and enterprise access coming at a later date.
ChatGPT also introduces a dedicated image home page, which allows users to generate images by describing what they want to display in their messages, and provides dozens of preset filters and prompts, as well as prompt hints for generating images.
The Verge, an IT news site, points out that OpenAI sees this new model as a shift from mere novelty to a practical, high-definition visual creation tool, aiming to turn ChatGPT into a 'creative studio in your pocket.' With competition in the image generation AI field intensifying due to the success of Google's Nano Banana image generation AI series, OpenAI appears to be aiming to appeal to the practicality of ChatGPT, particularly for businesses, in response to pressure from investors to monetize the technology.
Related Posts:
in AI, Web Service, Design, Posted by log1i_yk