OpenAI Launches ‘Images 2.0’ in ChatGPT with Smarter, More Realistic Visuals

New Delhi — OpenAI has introduced ‘Images 2.0’, a next-generation image generation model now integrated into ChatGPT, bringing notable improvements in realism, accuracy, and reasoning.

The upgraded model is designed to better interpret detailed prompts, place objects with higher precision, and handle complex elements like dense text, user interfaces, and multilingual content. It also supports flexible aspect ratios, making it suitable for everything from social media creatives to presentation visuals.

One of the key enhancements is the addition of “thinking” capabilities. When activated, the system can pull in real-time information via web search, generate multiple variations from a single prompt, and cross-check outputs for consistency and accuracy.

According to the company, this upgrade significantly reduces the effort needed to turn ideas into finished visual assets.

The model also performs more effectively across languages, especially in rendering non-Latin scripts such as Hindi, Japanese, Chinese, Korean, and Bengali—broadening its usability for global audiences.

In terms of output quality, ‘Images 2.0’ delivers more lifelike and stylistically accurate visuals across formats like photography, cinematic stills, manga, and pixel art, with improved handling of lighting, textures, and fine details.

OpenAI highlighted a wide range of applications, including UI mockups, magazine layouts, infographics, handwritten notes, comics, advertisements, and cinematic scenes. It also supports design workflows on platforms like Canva, Figma, and Adobe.

Developers can access the model via the ‘gpt-image-2’ API, enabling integration into products for use cases such as marketing, education, and content creation. The tool is also available within ChatGPT and Codex.

While the update marks a major leap forward, OpenAI acknowledged that challenges remain—particularly with highly complex spatial compositions and intricate repetitive patterns. Certain outputs, like detailed diagrams, may still require human verification.

To address safety concerns, the company has implemented multiple safeguards, including prompt and image-level moderation, along with provenance tools like metadata tagging and watermarking.

The new model is now available, with advanced capabilities offered to paid users. API access is also open, with pricing varying based on image quality and resolution.

With inputs from IANS