OpenAI has officially launched ChatGPT Images 2.5, an update designed to address some of the most persistent frustrations in AI image editing. Rolling out to users and developers, this new version promises greater precision, better detail preservation, and faster generation capabilities, alongside innovative creative tools.
Many users have experienced the challenge of asking an AI image generator to modify a background, only to find that crucial elements like faces or clothing are unintentionally altered. Similarly, making multiple successive edits often leads to a gradual degradation of the original image's integrity. OpenAI states that Images 2.5 is specifically engineered to overcome these hurdles.
Enhanced Precision and Detail Preservation
At the core of ChatGPT Images 2.5 is its focus on more precise editing. The model can now make targeted changes to an image without unnecessarily rebuilding the entire visual. This means if you wish to swap a product in a photograph, change a background, or update text, the AI is designed to leave other elements untouched.
This precision is particularly beneficial when working with images containing people or branded content. For instance, when placing a person into a new setting, the updated model aims to retain their recognizable features, preventing the subject from changing along with the background.
Beyond precision, Images 2.5 also boasts improved consistency across multiple edits. Users can now make successive modifications to an image without compromising earlier changes or seeing the overall quality diminish. OpenAI also reports up to a 50% reduction in image-generation latency compared to its predecessor, Images 2.0, coupled with more natural lighting and richer textures.
New Creative Tools for Users
Recognizing that not everyone can articulate visual ideas purely through text, OpenAI is introducing the Sketch tool. This feature allows users to draw directly within ChatGPT, using their sketch as a foundational reference for image generation. Imagine outlining a room layout, sketching an outfit concept, or drawing a rough poster design; ChatGPT can then combine this visual input with text instructions to produce a more refined image. Users can activate this by typing @Sketch.
The update also includes new templates for common formats like posters and merchandise designs, offering a starting point for creators. Additionally, users can now place comments directly on images for more focused editing requests and share the prompts used to create an image, fostering collaboration and creativity.
Expanding Capabilities for Developers
OpenAI is extending these advancements to developers with two new API models: GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst. Flare is tailored for general-purpose image generation, ideal for social media content, creator tools, and product experiences. Sunburst, on the other hand, is designed for more intricate creative workflows that demand exceptional precision across edits, such as professional campaign visuals and polished product imagery.
Availability
ChatGPT Images 2.5 is currently rolling out to all ChatGPT users, including those on ChatGPT Work and Codex, across desktop, mobile, and web platforms. Developers can access the GPT-Image-2.5 Flare and Sunburst models via the API.
This update marks a significant step towards making AI image editing more intuitive, reliable, and powerful for both casual users and professional creators, promising to reduce common frustrations and unlock new creative possibilities.