- OpenAI released ChatGPT Images 2.5, which features sharper details, faster generation, and smarter edits compared to its predecessor.
- The new model reduces image generation latency by as much as 50% compared to Images 2.0.
- Adobe has integrated OpenAI's image models into its Firefly creative AI studio.
- The Sketch feature allows users to draw directly in ChatGPT, which can then be used as a reference for image generation.
- Users can activate the Sketch feature by typing @Sketch into the chat box, allowing them to create drawings that ChatGPT can transform into images.
OpenAI's ChatGPT Images 2.5 introduces significant advancements in image generation, including sharper details and faster processing times. The model can now generate images with more natural lighting and richer textures, while also improving editing capabilities by following instructions across multiple turns.1
The new Sketch feature allows users to create doodles that can be transformed into detailed images, enhancing user interaction. Users can activate this feature by typing @Sketch in the chat box, enabling them to draw directly within ChatGPT. This innovation has been tested successfully, with users reporting impressive results from simple sketches.45
OpenAI has also released two versions of Images 2.5: GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst. Flare is designed for general use with lower latency, while Sunburst caters to premium workflows requiring tighter control over edits. According to OpenAI, over 3 billion images are generated weekly across its platforms, showcasing the model's widespread adoption.
The integration of these models into Adobe Firefly signifies a collaboration that combines leading AI technologies with Adobe's creative tools, enhancing the overall user experience in digital content creation. OpenAI's commitment to improving image generation is evident, as they continue to innovate in the AI space.3
“The update cuts image generation latency by up to 50% versus Images 2.0 and introduces two developer models, Flare and Sunburst, with Flare running two to four times faster in Manus's tests. OpenAI also reports over 3 billion images are created weekly across its image tools.”
