OpenAI launches ChatGPT Images 2.0 with full-page graphics and brand

ChatGPT Images 2.0, now live for all ChatGPT users, can generate full‑page graphics that include accurate text and match brand styles such as ZDNET’s visual identity, a leap from the “decorations” of earlier releases [1]. The upgrade matters because it turns a novelty tool into a practical asset for marketers and designers who need on‑the‑fly visuals without manual editing.

Expanded capabilities and real‑world performance

OpenAI’s new image engine moves beyond simple decorative outputs to produce complete graphics that embed contextual text. In a ZDNET hands‑on test, the model reproduced the ZDNET logo perfectly, matched the outlet’s color palette, and rendered tiny angled text without errors [1]. The same engine also generated sketchnotes of the US Bill of Rights in ZDNET’s style, delivering flawless wording—something Google’s Nano Banana struggled with in earlier comparisons [2].

Quantitatively, Images 2.0 scored 97 % in a suite of benchmark tests, outpacing Gemini’s Nano Banana (85 %) [2]. In a “admiral photo re‑contextualization” task, the model earned 14 out of 15 points, notably preserving the original face—a shortfall for Nano Banana, which altered facial features [2]. It also achieved a perfect 15/15 on a black‑and‑white restoration challenge, matching Nano Banana’s performance but with fewer visual artefacts [2].

Errors and the need for human oversight

Despite the gains, Images 2.0 still produces occasional factual slips. In an infographic about AI website builders, the model mis‑labelled the number of reviewed services (nine vs. five) and substituted an unreviewed competitor (Durable) for a listed vendor (10Web) [1]. Star‑rating tables it generated were also invented, diverging from the source text. OpenAI acknowledges that human review remains essential to catch such inaccuracies [1].

Competitive picture

The leap in text handling narrows the gap with Google’s Gemini platform, which previously excelled in basic image generation but faltered on prompt discipline and privacy‑related personalization [2]. OpenAI’s broader availability—Images 2.0 is accessible to free and paid ChatGPT users, though the most advanced language features still require a Plus subscription with the “Thinking” model enabled—gives it a distribution advantage over Gemini’s more limited rollout [1].

What to watch

• Feature rollout: OpenAI’s roadmap for extending brand‑style controls and reducing factual errors in Images 2.0.
• Google response: Updates to Gemini Nano Banana that address text accuracy and prompt consistency.
• Enterprise adoption: Early uptake of Images 2.0 in marketing and design teams, especially for rapid infographic creation.

ChatGPT Images 2.0 demonstrates that AI‑generated visuals are moving from novelty to utility, but the persistence of subtle errors means human oversight will remain a critical part of the workflow. The next few months will reveal whether OpenAI can tighten accuracy enough to make the tool a reliable staple for content creators.