OpenAI’s new image model, Images 2.0, is drawing attention from both the tech community and everyday users in Vietnam. For the first time, text embedded in AI-generated images is rendered with near-perfect accuracy- even in Vietnamese - fixing a long-standing weakness of earlier AI systems.

Just two years ago, tools like DALL-E 3 often produced misspelled words or even invented nonsensical terms when asked to generate images containing text. Now, Images 2.0 can create a complete restaurant menu with clear, natural content that is almost indistinguishable from human design.

In real-world testing with Vietnamese, ChatGPT integrated with Images 2.0 proved faster and more precise than the previous Image 1.5 version released late last year. Long passages of Vietnamese text now appear without spelling or character errors. Users can also choose from multiple aspect ratios, including square, portrait, landscape, and widescreen.

Community feedback suggests that Images 2.0 is putting significant competitive pressure on rivals, including Google’s Nano Banana Pro. One social media user joked: “After Google launched Nano Banana Pro, ChatGPT finally joined the game with Image 2.0. It understands Vietnamese better, with more detailed command structures. Looks like freelance menu designers will be busy taking orders.”

Although OpenAI has not disclosed full technical details, the company revealed that Images 2.0 is equipped with “reasoning” capabilities - able to self-check results, search for information, and generate multiple variations from a single request. This allows the model to design complex products such as marketing materials, user interfaces, or multi-panel comics with high levels of detail.

Support for non-Latin scripts like Japanese, Korean, and Hindi has also improved significantly. Output resolution can reach 2K, enabling fine reproduction of small details, symbols, and dense layouts - areas where earlier AI models struggled.

Images 2.0 is now available to all ChatGPT and Codex users. Paid subscribers gain access to advanced options, while OpenAI has also launched the gpt-image-2 API, with pricing based on output quality and resolution.

With these improvements, spelling errors in AI-generated images are virtually eliminated, opening the door to practical applications in design and communications.