I got an early look at ChatGPT Images 2.0, and it's impressive - with one exception
The new model acts as a 'visual thought partner' but can't accurately reproduce specific brand logos in early tests.
OpenAI has announced ChatGPT Images 2.0, a significant upgrade to its image generation model that reframes images as a 'visual language' rather than mere decorations. The model introduces enhanced 'thinking' capabilities, allowing it to process vague prompts, gather external data (like weather), and generate coherent, multi-image outputs such as infographics. It supports extreme aspect ratios from 3:1 to 1:3 and can render small text and UI elements at up to 2K resolution, offering improved precision and design control for complex compositions.
In an exclusive preview test for ZDNET, the model demonstrated its strength by creating a detailed 16:9 infographic about its own update, using brand guidelines from a provided homepage screenshot. However, it consistently failed to accurately reproduce the ZDNET logo, first rendering a 'droopy' Z and then, bizarrely, generating a pre-2022 version of the logo with the current color scheme. This highlights a critical weakness in brand fidelity that persists despite specific user instructions, indicating that while the model excels at conceptual visual tasks, precise logo replication remains a challenge.
- Introduces 'thinking' capabilities to act as a visual thought partner, building infographics from vague prompts by gathering external data.
- Supports extreme aspect ratios (3:1 to 1:3) and 2K resolution for detailed text, UI elements, and complex compositions.
- Early testing shows impressive infographic generation but persistent failure in accurately reproducing specific brand logos like ZDNET's.
Why It Matters
This advances AI from simple image generation to a collaborative design tool, though brand consistency for professional use remains a hurdle.