Stay updated with the latest news and exclusive content on AI advancements by subscribing to our daily and weekly newsletters. Find out more
Businesses now have the opportunity to integrate Studio Ghibli-inspired images created by ChatGPT into their operations. The model powering the popular image generation tool used in ChatGPT has been added to OpenAI’s API.
The gpt-image-1 model enables developers and companies to seamlessly incorporate high-quality, professional-grade image generation directly into their tools and platforms.
“The model’s versatility allows for the creation of images in various styles, adherence to custom guidelines, utilization of world knowledge, and accurate rendering of text — offering numerous practical applications across various industries,” as stated in a blog post by OpenAI.
Pricing for the API is divided into tokens for text and images. Prompt text input tokens are priced at $5 per 1 million tokens. Image input tokens cost $10 per million tokens, while generated image output tokens are priced at $40 per million tokens.
Competitors such as Stability AI provide a credit-based system for their API, where one credit equals $0.01. Utilizing their flagship Stable Image Ultra incurs a cost of eight credits per generation. Google’s image generation model, Imagen, charges users $0.03 per image generated through the Gemini API.
Centralized Image Generation
In April, OpenAI introduced the ability for ChatGPT users to generate and edit images directly within the chat interface, following the integration of image generation into ChatGPT through the GPT-4o model.
The company reported that image generation within the chat platform quickly became a popular feature, with over 130 million users accessing it and creating 700 million images within the first week alone.
However, this surge in popularity posed challenges for OpenAI. Social media users discovered they could prompt ChatGPT to create images inspired by Studio Ghibli, leading to a flood of similar photos across social media platforms. This trend prompted OpenAI CEO Sam Altman to mention that the company’s GPUs were under strain.
Prior to this, OpenAI had integrated its image model DALL-E 3 into ChatGPT. Unlike the native multimodal understanding of GPT-4o, DALL-E 3 was a diffusion transformer model.
Enterprise Applications
Enterprises are seeking the capability to generate images for their projects without the need for separate applications. By including the image model in its API, OpenAI enables enterprises to connect gpt-image-1 to their existing ecosystems.
OpenAI highlighted that several enterprises and startups have already utilized the model for creative projects, products, and experiences, mentioning well-known brands in their blog post.
Canva is exploring ways to incorporate gpt-image-1 into its Canva AI and Magic Studio Tools. GoDaddy has started experimenting with image generation for customers to design their logos, and Airtable now facilitates enterprise marketing and creative teams in managing asset workflows efficiently.
OpenAI assured that gpt-image-1 will have the same safety measures on the API as in ChatGPT. Images generated using the model will include metadata from the Coalition for Content Provenance and Authenticity (C2PA) to identify content as AI-generated and track ownership. OpenAI is a member of C2PA’s steering committee.
Users have the ability to moderate content to generate images that align best with their brand. OpenAI also committed not to utilize customer API data, including any images uploaded or generated by gpt-image-1, for training its models.
