The technology behind OpenAI’s improved image generator is now generally available to developers.
OpenAI made the technology available on Wednesday through the ChatGPT API. Developers can now integrate it into their applications, OpenAI writes in an announcement.
Popular Technology
The improved image generation went viral and more than 130 million ChatGPT users generated 700 million images in the first week alone. This put severe pressure on server capacity, and a limit was quickly imposed.
The image generation is powered by a multimodal AI model called ‘gpt-image-1’. This model is now available to developers via the API. They can generate multiple images simultaneously and determine the quality and speed. OpenAI says the same safety measures are applied as with ChatGPT. The difference is that developers can set the moderation sensitivity: ‘auto’ for standard filtering or ‘low’ for fewer restrictions.
For text, you pay five dollars per million input tokens. If you share reference images, it costs ten dollars per million input tokens, and for generated images, you pay 40 dollars per million input tokens. This amounts to 2 cents (low quality), 7 cents (medium quality), and 19 cents (high quality) per generated square image.
According to OpenAI, companies such as Adobe, Instacart, Canva, Figma, and Wix are already experimenting with the technology. Users of the Figma Design platform can generate and edit images, while Instacart is testing the model for photos of recipes and shopping lists.
read also