ChatGPT reduces design workflow to one
- OpenAI folded image generation and editing directly into ChatGPT and its API, letting users create, revise, and reuse images inside one conversation instead of bouncing between separate design tools. - OpenAI said the new ChatGPT Images model delivers more precise edits, more consistent details, and image generation up to four times faster, while the API supports multi-image editing in one request. - The shift turns image work into a chat workflow inside ChatGPT, with saved image libraries and in-thread edits replacing older prompt-export-edit loops. (openai.com)
OpenAI moved image generation and editing into ChatGPT itself, turning what used to be a multi-app design loop into a single chat. (openai.com) Image generation is the part of artificial intelligence that turns written instructions into pictures. OpenAI said its latest system is built into GPT-4o, so the same model can handle text, images, and follow-up edits in one conversation. (openai.com) That changes the mechanics of the job. Instead of drafting a prompt in one tool, exporting an image, opening an editor, and rewriting instructions, ChatGPT now lets users generate an image, mark an area, and ask for a change in the same thread. (help.openai.com) (developers.openai.com) OpenAI’s help center says every image created in ChatGPT is saved under an Images library on web and mobile, and users can reopen those files and keep editing them. The editor supports both direct conversational edits and targeted edits with a selection tool. (help.openai.com) OpenAI said in December 2025 that the upgraded ChatGPT Images model was rolling out to all users and would generate images up to four times faster, with more precise edits and more consistent details. The same release said the model was also available in the application programming interface as GPT Image 1.5. (openai.com) For developers, the same shift shows up in the application programming interface. OpenAI’s image generation guide says GPT Image models can generate images inside the Responses API, and the image edit endpoint accepts up to 16 source images for a single edit request. (developers.openai.com 1) (developers.openai.com 2) Microsoft, which offers OpenAI models through Azure AI Foundry, published benchmark results for GPT Image 1.5 showing 91.2% prompt alignment and 96.9% on diagram and flowchart tasks in its internal dataset. Microsoft also said the model performed best on single-turn modification. (techcommunity.microsoft.com) Those numbers do not match the “93% win rate” claim circulating in creator commentary, and OpenAI’s public materials reviewed here do not publish that figure. The documented change is more concrete: image creation, revision, storage, and reuse now happen inside ChatGPT and the same underlying application programming interfaces. (openai.com) (help.openai.com) (developers.openai.com) The practical result is that design work starts to look less like file shuffling and more like versioning in a chat. OpenAI’s product pages describe the system as a native image capability of GPT-4o, and the product flow now reflects that claim. (openai.com)
