Last month, OpenAI released a major update to ChatGPT’s image generation feature, powered by a new model. The update took the world by storm, with more than 130 million users generating more than 700 million images in just one week of its launch.
Today, OpenAI is making this improved image generation model available to developers via the gpt-image-1 API. The new API enables applications and services to integrate high-quality image generation, support for a wide range of visual styles, precise image editing, rich world knowledge, and sharp text rendering.
gpt-image-1 is now available on the Images API, with plans to support the Responses API soon.
Flexible control features:
- Moderation : Developers can adjust moderation sensitivity via the moderation parameter. By default, the standard filter is applied, but low mode can be selected to reduce restrictions.
- Customize output : Control image quality, number of images generated, transparent/opaque background, and output format (JPEG, PNG, WebP).
This is a paid subscription, and the cost is token-based with specific milestones as follows:
- Text input : 5 USD/1 million tokens.
- Image input : 10 USD/1 million tokens.
- Image output : $40/1 million tokens.
Real-world example: Each low, medium, and high quality square image costs about $0.02, $0.04, and $0.19, respectively.
Developers can try out this model on the OpenAI Playground. Many companies such as Adobe, Airtable, Figma, Gamma, HeyGen, Wix, Photoroom, Canva, GoDaddy, HubSpot, Instacart, Invideo, and OpusClip have integrated OpenAI's photo generation API into their products and received positive feedback from the user community.