OpenAI announced the release of 'ChatGPT Image 2.5' on September 9, which enhances image quality and editing accuracy.
According to OpenAI, Image 2.5 accurately reflects the unique characteristics of people and objects while naturally expressing lighting and texture. Users can depict individuals from reference photos in new backgrounds or artistic styles, or change the shooting background and composition while maintaining the product's appearance. The ability to modify only the requested parts of complex images while preserving other details has also improved.
Consistency in iterative editing has been enhanced. When refining product photos or advertising materials step by step, the occurrence of changes to previous edits or a decline in image quality has been reduced.
A new sketch feature allows users to fine-tune the creative process. By entering '@Sketch' in the ChatGPT chat window, users can access sketch tools and draw directly with their mouse.
OpenAI also released two API models for developers: 'GPT-Image-2.5 Flare' and 'GPT-Image-2.5 Sunburst.' The Flare model offers higher image quality than GPT-Image-2 while reducing generation latency by up to 50%.
Sunburst is designed for tasks requiring detailed adjustments and high precision during the editing process, even if it takes longer to generate.
OpenAI is gradually providing Image 2.5 to users across all subscription plans.
Meanwhile, OpenAI Korea held a press conference to mark its first anniversary and discuss its direction in the South Korean market.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.