OpenAI Introduces ChatGPT Images 2.5 Model

OpenAI has introduced the new ChatGPT Images 2.5 image generation model. The new version is up to 50% faster than the previous Images 2.0 model. It also renders natural light and textures more accurately and better preserves key objects from uploaded reference images.

The new model has also expanded image editing capabilities. Users can now modify specific areas not only by providing text descriptions but also by leaving annotations directly on the image or highlighting the desired object.

A Sketch feature has also been added to ChatGPT. It allows users to draw their ideas directly and use them as a reference for the final image.

Ready-made templates for popular formats such as posters and product images have also been introduced. This allows users to work from a pre-defined layout instead of starting from scratch.

OpenAI states that Images 2.5 maintains key image aspects better during multiple consecutive editing steps. For example, when an object or background is changed, other details are more likely to remain consistent.

Two new models have also been released for developers via API. GPT-Image-2.5 Flare combines speed, quality, and editing capabilities, designed for high-volume generation, social media content, and product imagery.

GPT-Image-2.5 Sunburst is designed for creative and editing tasks that require rich detail and precise control. Its generation time may be longer compared to Flare.

ChatGPT Images 2.5 is being rolled out to users of ChatGPT on desktop, mobile, and web, as well as ChatGPT Work and Codex.