OpenAI rolls out ChatGPT Images 2.0 with better text in images

In short: OpenAI is releasing ChatGPT Images 2.0, a new image maker inside ChatGPT that is better at putting readable text into images.

What happened

OpenAI announced a new model called ChatGPT Images 2.0. It is designed to create images from written prompts, like “make a Mexican restaurant menu” or “draw a comic strip.”

One standout change is that it can produce clearer, correctly spelled text inside images. In the past, many AI image tools struggled with words on signs, menus, and buttons, often making letters look scrambled. The TechCrunch example shows Images 2.0 generating a menu that looks usable, instead of nonsense words.

OpenAI says the model can also handle non-Latin writing better, including Japanese, Korean, Hindi, and Bengali. The company did not say what exact type of image model it uses. OpenAI also said the model has “thinking capabilities,” meaning it can do extra steps like check its own work and create multiple versions from one request (like making the same ad in different sizes).

All ChatGPT and Codex users are expected to get access starting Tuesday. Paid users will be able to generate more advanced results. OpenAI also plans to offer the same image tool through an API (an API is like a plug that lets other apps connect to the tool), with prices depending on image quality and size.

Why it matters

Text inside images is everywhere, like menus, posters, app screens, and product labels. If AI tools can reliably create those without obvious mistakes, it becomes easier for small businesses and creators to make basic marketing materials quickly. It could also make it harder for everyday people to spot what is AI-made just by looking for misspellings.

Source: TechCrunch AI