OpenAI Rolls Out ChatGPT Images 2.5 Across ChatGPT, Work and Codex

OpenAI today launched ChatGPT Images 2.5, an upgraded image-generation model that improves detail, editing accuracy and generation speed while adding new creative tools inside ChatGPT. The model is rolling out across ChatGPT, ChatGPT Work and Codex on desktop, mobile and web, while developers are getting two API versions: GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst.

OpenAI says Images 2.5 can generate more natural-looking lighting and textures while doing a better job of keeping people and other subjects recognizable when working from reference photos. It is also designed to follow targeted editing requests more consistently, including across longer conversations where users make several changes to the same image.

Generation speed has improved as well. OpenAI says latency is up to 50% lower than with Images 2.0, allowing users to create and revise images more quickly.

The release also expands the controls available directly in ChatGPT. A new Sketch feature lets users draw inside ChatGPT and use the drawing as a visual reference for an image. Someone could sketch the shape of an outfit or the arrangement of a room, for example, and then provide additional instructions describing how the finished result should look.

Templates provide another starting point for image creation. Users can select formats such as posters or merchandise and then specify the information, visual elements and style they want included. OpenAI has also added the ability to place comments directly on images to identify specific areas for editing.

Prompt sharing is being expanded alongside those tools. When users share an image, they can choose to include the prompt behind it, allowing someone else to reuse the concept with different photos or other personalized details.

A major focus of Images 2.5 is maintaining consistency while an image is repeatedly edited. OpenAI says the model is better at changing only the requested element while preserving surrounding subjects, composition and other visual details. Earlier modifications are also more likely to remain intact as additional edits are made.

Those improvements are intended to carry over into developer workflows. OpenAI says applications using the Images API can make narrower changes to elements such as products, backgrounds or text without unnecessarily altering the rest of an asset.

Images 2.5 also expands the model's ability to interpret complex visual requests. OpenAI says it can produce more accurate real-world information inside images, handle more complicated layouts and work with transparent backgrounds. The model is also intended to stay closer to requested visual styles and compositions when prompts contain many individual requirements.

For API customers, GPT-Image-2.5 Flare is positioned as the default option for most applications. OpenAI says it provides higher image quality than GPT-Image-2 while cutting latency by 50%, with potential uses including social content, product experiences, visual search, rapid prototyping and high-volume image generation.

GPT-Image-2.5 Sunburst is aimed at workflows requiring more precise control over detailed edits. OpenAI points to finished campaign material and polished product imagery as examples of the type of work it is designed to support.

The company says people now create more than 3 billion images each week across ChatGPT Images and GPT-Image models accessed through the API. With Images 2.5, OpenAI is combining improvements to the underlying model with more hands-on tools for drawing, editing, templating and sharing images inside ChatGPT.

OpenAI said Images 2.5 continues to use safeguards that check prompts and generated images for harmful content. Images created with its tools also continue to include C2PA metadata and invisible watermarking.

See more here:

This analysis is based on reporting from OpenAI.

Image courtesy of OpenAI.

This article was generated with AI assistance and reviewed for accuracy and quality.