GPT Image 2 launch

OpenAI released ChatGPT Images 2.0 on April 21, adding sharper text rendering, multilingual output, and support for more complex image prompts. (openai.com) The new model can generate images with more accurate lettering, better backgrounds, and stronger visual reasoning, according to OpenAI’s product post. OpenAI said the system is designed to handle layouts, diagrams, posters, and other images where words inside the picture need to stay legible. (openai.com) PCMag reported on April 22 that the update also adds web access, multi-image outputs, and support for non-Latin text inside generated images. The outlet said OpenAI rolled the model out Tuesday as an upgrade to ChatGPT’s image generator. (pcmag.com) Image generators turn text instructions into pictures, but earlier systems often mangled signs, menus, labels, and comic dialogue. OpenAI said Images 2.0 improves “text rendering” and “advanced visual reasoning,” which means the model is better at placing words and objects where a prompt asks for them. (openai.com) That problem has mattered since OpenAI folded image generation into GPT-4o on March 25, 2025, instead of keeping it as a separate art tool. In that launch, OpenAI said it wanted image generation to be “useful,” not just visually striking, with an emphasis on precise prompts and real-world assets. (openai.com) The developer side moved in the same direction. OpenAI’s image-generation documentation says the Responses API supports image generation inside multi-step conversations, accepts image inputs, and now includes GPT Image models such as gpt-image-2 for generating or editing visuals. (developers.openai.com) OpenAI’s community forum said gpt-image-2 is available in the application programming interface and in Codex, with a focus on editing, layout accuracy, instruction-following, and production workflows. The post described the model as built for cases where images need to be “usable,” not just illustrative. (community.openai.com) The immediate pitch is less time fixing broken text after the image is made. For designers, marketers, and small businesses making mockups, menus, or signage, that shifts more of the work from cleanup in editing software to the first prompt. (pcmag.com) OpenAI’s release page frames Images 2.0 as the next step after last year’s GPT-4o image rollout: fewer garbled words, better multilingual support, and more reliable image construction from detailed instructions. The company is now treating image generation less like a novelty feature and more like a document-and-design tool inside ChatGPT. (openai.com)