The distinction between human-created and AI-generated images has blurred significantly with the introduction of ChatGPT's new Images 2.0 model. Just two years ago, AI struggled to produce coherent menus, often resulting in bizarre culinary terms. However, the latest version now generates realistic menus for Mexican cuisine that could easily pass for those crafted by human chefs, sparking curiosity about its accuracy.
Unlike its predecessors, which often faltered in spelling and coherence, the Images 2.0 model showcases a remarkable leap in quality. As noted by Asmelash Teka Hadgu, founder of Lesan AI, traditional diffusion models relied heavily on reconstructing images from noise, limiting their ability to accurately depict textual elements within images. This new model appears to have overcome these challenges, enabling it to create visually appealing and precise content.
Researchers have shifted focus towards autoregressive models for image generation, which predict image features in a manner similar to large language models. Despite inquiries, OpenAI has not disclosed the specific model architecture behind Images 2.0, but they have highlighted its enhanced capabilities.
OpenAI describes Images 2.0 as possessing advanced "thinking capabilities," allowing it to search the web, generate multiple images from a single prompt, and verify its outputs. This model is adept at producing marketing materials in various formats, including multi-paneled comic strips. It also demonstrates improved understanding of non-Latin scripts, including Japanese and Hindi, although its knowledge base is limited to information available until December 2025.
The press release emphasized that Images 2.0 achieves a new level of detail and specificity, effectively following instructions and rendering intricate elements such as small text and UI components at resolutions of up to 2K. While generating complex images may take a few minutes, the process remains significantly faster than traditional methods.
Starting Tuesday, all ChatGPT and Codex users will gain access to Images 2.0, with premium users enjoying the ability to create more sophisticated outputs. OpenAI will also provide an API for this model, with pricing based on output quality and resolution.
As technology continues to evolve, the implications of AI-generated imagery could reshape industries ranging from marketing to entertainment, paving the way for more innovative and creative applications in the future.
