OpenAI has introduced a new feature for its GPT-Image-2 model that allows users to generate images with transparent backgrounds through its API. This update enables the creation of PNG files without any background, offering improved visual quality for specific use cases. Users can activate the feature by including the 'background=transparent' parameter in their requests, according to OpenAI.
The feature is designed to enhance the accuracy of image generation, particularly for intricate elements such as transparent glass or thin fibers. OpenAI claims that integrating the alpha channel during the generation process yields better results than traditional background removal techniques. The model produces raster graphics and does not guarantee pixel-perfect accuracy. Users are advised to verify generated values manually, especially for charts with exact numerical data. Setup requires Python, the OpenAI and Pillow libraries, and an API key.
OpenAI highlighted the practical applications of the transparent background feature, including product shots for online stores, PowerPoint diagrams, design elements like icons and stickers, and merchandise artwork. The company emphasized that the model does not automatically include background descriptions in prompts, as it may still generate one otherwise. The update is part of a broader effort to refine the capabilities of GPT-Image-2 for diverse creative and professional tasks.
Source: thedecoder
