On Wednesday, OpenAI made its advanced image generation technology—previously introduced in ChatGPT—available to developers via its API. This allows developers to integrate the tool into their own applications and platforms.
The updated image generator, which rolled out to most ChatGPT users in late March, quickly gained popularity for producing lifelike visuals in styles like Studio Ghibli and AI-themed action figures. While the release sparked a surge in user growth, it also placed significant strain on OpenAI’s systems. In just the first week, over 130 million ChatGPT users created more than 700 million images.
Now available through the API, the tool is powered by the “gpt-image-1” model—a fully multimodal AI system. This model can produce images in various styles, follow detailed instructions, draw from real-world knowledge, and even generate text within visuals.
Developers using the API can request multiple images at once and choose between different quality settings, which affect both the image fidelity and generation speed.
OpenAI states that gpt-image-1 includes the same safety measures used in ChatGPT’s image generation. These include content filters that prevent policy-violating outputs. Developers can adjust how strictly the model moderates content, with options for “auto” (standard filtering) and “low” (less restrictive filtering). The “low” setting allows for broader content generation, though still within defined limits.
All images created by the model include a watermark via C2PA metadata, ensuring they can be identified as AI-generated by compatible systems and platforms.
Pricing is based on token usage: $5 per million input tokens for text, $10 per million for images, and $40 per million output tokens for image results. Based on OpenAI’s estimates, generating a square image costs approximately 2 cents for low quality, 7 cents for medium, and 19 cents for high quality.
Several companies, including Adobe, Airtable, Wix, Instacart, GoDaddy, Canva, and Figma, are already using or testing gpt-image-1. For instance, Figma now supports image generation and editing through the model, while Instacart is exploring its use for generating recipe-related visuals.
