Photo credit: venturebeat.com
Businesses can now seamlessly integrate image generation inspired by Studio Ghibli into their workflows, thanks to a new tool from OpenAI. The company has made available the model behind its acclaimed image generation technology, which is already being utilized in ChatGPT, for widespread API use.
The gpt-image-1 model empowers developers and organizations to “embed high-quality, professional-grade image generation directly into their own applications and platforms.”
According to OpenAI, “The model’s flexibility enables it to produce images in various styles, adhere to specific guidelines, utilize global knowledge, and accurately depict text — paving the way for numerous practical applications across different fields,” as stated in a recent blog post.
The pricing structure for the API distinguishes between text and image tokens. Text input tokens will cost $5 for every million tokens, while image input tokens are priced at $10 per million tokens. The cost of image output tokens, reflecting the generated images, stands at $40 for each million tokens.
In comparison, rivals like Stability AI have adopted a credit-based pricing model, where one credit is equivalent to $0.01. Each image generation using its flagship Stable Image Ultra requires eight credits. Google’s Imagen model charges users $0.03 for each generated image through the Gemini API, as outlined in their pricing documentation.
Centralized Image Generation
This development follows OpenAI’s initiative earlier this year to enable ChatGPT users to produce and modify images directly within the chat interface after introducing image generation via the GPT-4o model.
OpenAI reported that this feature swiftly gained popularity among users, with the platform recording over 130 million interactions and 700 million images created in just its first week.
This surge in interest, however, brought certain challenges. Social media users quickly began using ChatGPT to generate Ghibli-inspired images, leading to a flood of similar visuals across platforms. This phenomenon prompted OpenAI’s CEO, Sam Altman, to humorously mention that their GPUs “are melting.”
Before this, OpenAI had incorporated its image model DALL-E 3 into ChatGPT, which operates on a different principle, using a diffusion transformer model compared to the multimodal capabilities of GPT-4o.
Applications in Enterprise
Organizations are increasingly seeking ways to generate images tailored to their projects without the necessity of switching between applications. By adding this image model to its API, OpenAI provides businesses with the ability to integrate gpt-image-1 directly into their existing systems.
OpenAI has noted that various enterprises and startups have already begun utilizing the model for creative endeavors, highlighting its application across notable brands in their communications.
For instance, Canva is reportedly assessing how to incorporate gpt-image-1 into its Canva AI and Magic Studio Tools. Meanwhile, GoDaddy is experimenting with image generation features to aid customers in logo creation, and Airtable is streamlining asset workflow management for its enterprise marketing teams.
Furthermore, OpenAI assures users that gpt-image-1 will incorporate the same safety measures as found in ChatGPT, including the embedding of metadata from the Coalition for Content Provenance and Authenticity (C2PA) to identify AI-generated content and track ownership. OpenAI is an active participant in C2PA’s steering committee.
Additionally, clients will have the capability to manage content moderation effectively, ensuring that the generated images align with their brand identity.
Importantly, OpenAI has committed that it will not utilize any client API data—such as images uploaded or produced by gpt-image-1—to enhance its models in the future.
Source
venturebeat.com