OpenAI Launches ChatGPT Images 2.5, Boosting Generation Speed by 50%

OpenAI on Tuesday announced the official launch of ChatGPT Images 2.5, a significant upgrade to its generative AI tool that can create images up to 50% faster than the previous version. The company also revealed the scale of its platform's adoption, stating that users now generate over 3 billion images weekly through ChatGPT and its API.

Key Upgrades and Features

According to OpenAI, the new model delivers substantial improvements in image quality, editing precision, and performance. The company highlighted that Images 2.5 produces more natural lighting and richer texture detail. It also demonstrates better adherence to user instructions during multi-step editing sessions and is more effective at preserving the key features of a subject from a reference photo.

New user-facing features have been rolled out across desktop, mobile, and web versions for all ChatGPT, ChatGPT Work, and Codex users. These include:

• Sketch: Allows users to provide a hand-drawn sketch directly within ChatGPT as a reference for the final image.
• Templates: Offers pre-configured starting points for common image formats like flyers and product photos.
• In-image annotations: Users can now add notes directly onto an image to guide specific edits.

New API Models for Developers

Ad

For developers using the API, OpenAI has released two distinct models to cater to different needs. The first, `GPT-Image-2.5 Flare`, is designed to match the quality, speed, and editing performance of the model available in ChatGPT.

The second model, `GPT-Image-2.5 Sunburst`, is engineered for higher precision and is intended for detailed creative work, though it comes with a longer generation time. Both models feature improved handling of reference photos, maintaining the subject's identity across various scenes and styles.

Safety and Context

OpenAI stated that this release builds upon its existing safety mechanisms with enhanced reviews of both text prompts and image outputs to prevent the generation of harmful content. The company reaffirmed its commitment to transparency and provenance by continuing to use C2PA metadata and invisible watermarking to identify images created with its tools.

The update underscores the rapid pace of development in the generative AI space, as major technology firms compete to offer more powerful and efficient creative tools. The significant speed improvement and new editing features are aimed at both casual users and professional creators who rely on the platform for a wide range of applications.