OpenAI recently unveiled ChatGPT Images 2.0, its latest image generation engine, which introduces a significant leap in functionality. Moving beyond creating simple "decorations," the new engine is now capable of producing full-page graphics that incorporate detailed and accurate text, signaling its readiness for more complex professional applications.
A tech editor, who engaged with an early pre-release version of Images 2.0, observed initial inconsistencies, particularly in accurately rendering specific brand elements like the ZDNET logo. However, an extensive hands-on evaluation conducted after the official release, utilizing a ChatGPT Plus account with the advanced "Thinking" model enabled, revealed substantial improvements across a diverse set of challenges.
A key aspect of the testing methodology involved optimizing the input for brand consistency. Rather than relying on the AI to autonomously identify and extract logos from uploaded web pages, the editor proactively provided a standalone ZDNET logo image alongside each prompt. This targeted approach proved instrumental in significantly improving the AI's ability to integrate brand identity correctly. Furthermore, it's important to note the operational context: due to ZDNET’s existing data access restrictions concerning OpenAI, the article content used for testing was supplied to ChatGPT via full-screen screenshots, captured using a Chrome extension, enabling the AI to "read" the textual information.
The first practical test focused on brand logo preservation and stylistic adherence. The prompt given to ChatGPT Images 2.0 was explicit: "Create a detailed and vivid infographic of this article using the ZDNET brand style and the attached ZDNET logo." The outcomes were notably successful. The generated infographic not only depicted the ZDNET logo correctly but also perfectly replicated the brand's specific color palette. Critically, the image distinguished itself through its text accuracy: all textual elements, even fine print angled within the graphic, were rendered without error, showcasing a robust capability in maintaining textual integrity within complex visual outputs.
The second challenge extended to generating styled "sketchnotes." This test recalled a prior experience with Google's Nano Banana, which, despite producing visually appealing images, struggled repeatedly to ensure correct wording in its sketchnote outputs. For ChatGPT Images 2.0, the stakes were intentionally raised. The editor requested sketchnotes of the US Bill of Rights, specifically demanding adherence to ZDNET's unique branding style. This particular focus on integrating detailed, critical text with precise visual branding aims to rigorously assess the AI's capacity to deliver high-value assets in professional design and content creation workflows.