OpenAI announced Wednesday that it has released its enhanced image generation technology, previously introduced in its ChatGPT platform, to developers via the company’s API. This move enables third-party app makers and service providers to incorporate advanced AI-generated imagery into their offerings.
Initially rolled out to most ChatGPT users in late March, OpenAI’s image generator quickly captured widespread attention due to its striking ability to produce realistic images, notably in Studio Ghibli’s distinctive artistic style and creative AI-generated figurines. The impact was immediate but brought with it significant operational challenges, as OpenAI reported over 130 million users generating around 700 million images within just one week of its public release—stretching the company’s computing capacities to their limits.
The AI technology powering the newly accessible API integration is a multimodal model dubbed “gpt-image-1.” This model can proficiently produce images spanning a variety of styles, adhere closely to specific user instructions, effectively utilize existing knowledge from the outside world, and accurately render text within image outputs. Developers leveraging the API can swiftly generate multiple images simultaneously and tailor the balance between image quality and generation speed to meet their specific needs.
OpenAI has assured developers that the same safety standards applied to ChatGPT’s image generation will be upheld in API usage. The safeguards prevent the generation of images that violate OpenAI’s guidelines, and developers have the flexibility to select moderation levels—automatic for strict, standard filtering, or “low” for less restrictive oversight. The “low” moderation option filters fewer categories of potentially sensitive content, according to OpenAI.
Additionally, all images produced by gpt-image-1 are embedded with C2PA metadata watermarks, a technical move designed to transparently identify images as AI-generated content across supported platforms and services.
Pricing for the API is structured around a token-based approach, with charges set at $5 per million input tokens for text prompts, $10 per million input tokens for image prompts, and $40 per million output tokens for high-quality image results. In practical terms, this equates roughly to costs of about 2 cents per low-quality, 7 cents per medium-quality, and 19 cents per high-quality square image creation, per OpenAI’s documentation.
Several major organizations have already incorporated or begun trials of the gpt-image-1 model in their workflows and products. Notable names such as Adobe, Airtable, Wix, Instacart, GoDaddy, Canva, and Figma are listed among early adopters. Figma has enhanced its design application suite by integrating gpt-image-1 to offer real-time image generation and editing features, whereas Instacart is currently evaluating the model for use in dynamically created recipe and product images.
