ChatGPT Images 2.5 is here: OpenAI wants to fix one of AI image editing’s biggest frustrations

Ever asked an AI image generator to change the background of a photograph, only to find that it has also altered your face, clothes or other details? Or tried making several edits to the same picture, only to watch the original image slowly lose its resemblance to what you started with? OpenAI says its latest image-generation update is designed to address these problems.

The company introduced ChatGPT Images 2.5, focusing on more precise editing, better detail preservation and faster image generation. Instead of simply producing a fresh image from a text prompt, the updated model is designed to make targeted changes while retaining the elements you want to keep.

MUST READ | Meta joins Microsoft and Google in offering voluntary exits to employees: Report

OpenAI says Images 2.5 can generate images with more natural lighting and richer textures, while following editing instructions more reliably across multiple rounds. Image-generation latency has also been reduced by up to 50% compared with Images 2.0.

Want to change just one thing in a photo? ChatGPT Images 2.5 aims to do that

One of the most useful changes is improved precision editing. You can ask ChatGPT to replace a product, change a background or update a piece of text without having the model unnecessarily rebuild the rest of the image.

This is particularly relevant when working with photographs of people, products or branded material. For instance, if you want to place yourself in a different setting, the model is designed to retain more of your recognisable features instead of changing the subject along with the background.

DON'T MISS | No more free Blinkit ambulance rides for all: Deepinder Goyal announces ₹2,500 charge

OpenAI also says the model maintains better consistency across multiple edits. This means you can make successive changes to an image without earlier edits becoming less consistent or the overall quality deteriorating.

For you as creators, this could make it easier to produce different versions of a visual. For everyday use, it could mean spending less time correcting details that an AI model changed without being asked.

ChatGPT gets a Sketch tool: Turn rough drawings into images

Not everyone can describe a visual idea in the right words. OpenAI is addressing that with Sketch, a feature that lets users draw directly in ChatGPT and use the drawing as a reference for image generation.

You could, for example, sketch the layout of a room, outline an outfit or draw a rough concept for a poster. ChatGPT can then use that drawing, along with your text instructions, to create a more finished visual. Users can start the feature by typing “@Sketch” in ChatGPT.

The update also introduces templates for formats such as posters and merchandise designs. Instead of starting with a blank prompt, you can select a format and customise it with their own text, design elements and style preferences.

Another addition allows you to place comments directly on images for more focused edits. You can also share the prompts used to create an image, so others can try the same idea with their own photographs and details.

OpenAI brings two new image models to developers

The update also includes two API models, the GPT-Image-2.5 Flare and the GPT-Image-2.5 Sunburst.

Flare is designed for general-purpose image generation, including social media content, creator tools and product experiences. Sunburst is aimed at more detailed creative workflows where greater precision across edits is important, such as campaign visuals and polished product imagery.

ChatGPT Images 2.5 availability

OpenAI says ChatGPT Images 2.5 is rolling out to all ChatGPT users, including ChatGPT Work and Codex users, across desktop, mobile and web. Developers can also access GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst through the API.