ChatGPT’s new version of Graphics 2.0 is amazing for creating audio

It used to be easy to distinguish human-generated images from AI – two years ago, you couldn’t use images create a menu for a Mexican restaurant without introducing new favorites like “enchuita,” “churiros,” “burrto,” and “margartas.”

Now, when I ask for the new version of ChatGPT Images 2.0 for Mexican food, it creates something that can be used in a restaurant without realizing that something is off. (However, the price of the ceviche at $13.50 might make me question the quality of the fish.)

For comparison, here are the results I got from DALL-E 3 two years ago (at that time, ChatGPT did not generate images):

AI image generators have them in history he struggled to write because they often use diffusion models, which work to reproduce images from noise.

“Media models (…) are redefining what is provided,” Asmelash Teka Hadgu, founder and CEO of Lesan AI, he told TechCrunch in 2024. “We can think of the text in the image as a very small area, so the image generator learns the features that cover many pixels.”

Researchers have been exploring other ways to create images, such as autoregressive modelswhich predicts how the image will look and function as an LLM.

Unfortunately, OpenAI refused to answer a question in a press release this week about which version supports ChatGPT Images 2.0.

The company, however, explained that the new model has “imagination,” which gives the power to search the Internet, create more images from the quick, and double-check what was created – this helps Images 2.0 to create commercial goods in different formats, as well as lines of multi-photographed images.

OpenAI also claims that Graphics can understand non-Latin words without translating them into languages ​​such as Japanese, Korean, Hindi, and Bengali. This type of notification expires in December 2025, which may affect how they can release other information related to current affairs.

“Images 2.0 brings unprecedented authenticity and integrity to image production.” It can not only imagine advanced images, but it brings the vision to life, able to follow the instructions, save the requested information, and provide visual elements that often break the images: small text, images, pictures, UI text and all UI elements. to the 2K concept,” OpenAI said in a press release.

This capability means that creating images isn’t as fast as writing a question in ChatGPT, but creating something more complex like a multi-faceted joke takes just a few minutes.

All ChatGPT and Codex users will have access to Images 2.0 starting Tuesday; paid users will be able to create high-quality output. The company will also develop gpt-image-2 API is availableprices depend on the quality and quality of the products.