OpenAI Introduces Direct Transparent Background Generation in GPT-Image-2

Release Overview

OpenAI has rolled out a new feature for its image generation model, "GPT-Image-2," enabling the direct generation of images without backgrounds. This eliminates the need for post-generation subject cutout work using external tools, realizing a streamlined workflow that is completed entirely within the generation process.

Details of the New Feature

With this update, users can instruct the model directly through prompts to generate "subjects without backgrounds." The resulting images are output with specific subjects placed on a transparent background, making them instantly ready for use in applications such as logo creation, presentation materials, and web design.

Technical Background

GPT-Image-2 enhances context control capabilities, allowing it to eliminate backgrounds during the model's inference process. This makes it possible to clearly preserve the outlines of subjects while outputting assets that include alpha channel (transparency) information.

Future Outlook

OpenAI remains committed to continuous model performance improvement, and this new background removal feature represents a significant step forward in enhancing the practical utility of generative AI in business environments. Further developments are expected to focus on increasing editing flexibility and prompt adherence.