OpenAI's GPT Image and DALL·E models generate and edit images from text prompts, available directly inside ChatGPT and through the OpenAI API. Built by OpenAI, the defining trait of its image tools is verifiable provenance: outputs carry C2PA metadata and, per OpenAI's own documentation, SynthID watermarking, while users retain full ownership of what they create under OpenAI's Terms of Use.
Who builds it
OpenAI develops GPT Image and DALL·E as part of the same product family that powers ChatGPT. Image generation is not a separate destination site: it is built into the ChatGPT interface, so anyone with a ChatGPT account can create or edit images inside an ongoing conversation. Developers who need image generation outside ChatGPT can call the same underlying models through the OpenAI API, using OpenAI's official image-generation guide as the reference implementation for integrating the models into other products.
What it does
GPT Image turns written prompts into new images and can also edit existing ones. OpenAI's platform documentation and help center both describe inpainting-style editing, where a user modifies part of an uploaded or previously generated image rather than starting over from a blank prompt. Generated images render at resolutions up to 1536×1024 pixels. Text rendering inside images is a supported capability — useful for posters, diagrams, product mockups or social graphics that need legible words rather than the garbled lettering earlier-generation image models were known for.
Pricing
OpenAI's image tools follow a freemium model. According to OpenAI's help center, ChatGPT users can generate images at no cost within usage limits before needing to upgrade. Broader, higher-limit access comes bundled into ChatGPT's paid plans, which start from $20/month according to OpenAI's pricing page. Developers who prefer to skip the ChatGPT interface can call the image-generation endpoints directly through the OpenAI API, billed separately under OpenAI's standard platform pricing rather than the ChatGPT subscription.
Rights and provenance
Ownership sits with the user. Under OpenAI's Terms of Use, you own the images you generate, and OpenAI's help center confirms the company does not claim copyright over outputs created via the API. Commercially, OpenAI treats generated images as fully usable: its Terms of Use describe full commercial-use rights for outputs, without the tiered, enterprise-only restrictions some rivals apply. For organizations with EU data-residency requirements, OpenAI's help documentation describes an optional EU data-residency setting that keeps inference within the EU — relevant for teams bound by GDPR-driven infrastructure rules. Every generated image also carries provenance signals: OpenAI's help center confirms that outputs include C2PA metadata, the open standard used across the industry to mark AI-generated content, together with SynthID watermarking. Combined, this means an image made with GPT Image or DALL·E arrives with both a clear rights answer — the user owns it and can use it commercially — and a clear provenance answer: it is machine-checkable as AI-generated, without requiring an enterprise-only indemnification scheme.