Back to all articles
•Technology

AI Advances in Image Generation: ChatGPT Images 2.0 Redefines Possibilities

View original source

The article highlights the significant progress in AI-generated imagery, focusing on the new ChatGPT Images 2.0 model by OpenAI.

  • AI-generated content has evolved rapidly; two years ago, AI struggled to create believable images for something as simple as a menu.
  • Asmelash Teka Hadgu, founder of Lesan AI, explains that diffusion models have been historically used, but new mechanisms such as autoregressive models offer improved predictions similar to large language models.
  • OpenAI hasn’t disclosed the exact nature of the model used, but states it has "thinking capabilities." This enables the creation of multiple images from one prompt and adjustment of marketing assets.
  • The new model also manages non-Latin text well in languages such as Japanese and Hindi as declared by OpenAI.
  • Despite its capabilities, the model's knowledge is current only up to December 2025, affecting the accuracy of more recent information.
  • With a resolution of up to 2K, Images 2.0 can produce complex visuals like comic strips within minutes.
  • The new model is available to all ChatGPT and Codex users with varying output levels based on subscription.