OpenAI may launch new image AI model soon, can offer more realistic images compared to rivals
Follow UsNew model aims to fix common AI flaws like unnatural lighting, textures, and facial details
Expected to improve composition with better text placement and structured visuals
ChatGPT may also get a “reference photo” feature for easier image generation workflows
Weeks after shutting down Sora, OpenAI is reportedly working on a new image generation model which can offer more realistic images compared to the rivals. The upcoming model is said to be called as gpt-image-2 and is said to be built on the recent surge in popularity driven by ChatGPT’s image tools, which went viral for creating stylised visuals such as anime-inspired artwork.
SurveyAs per the reports, the new model will focus on realism and aim to reduce that ‘artificial’ look which we often see in the AI generated images. Previous outputs were frequently criticised for overly smooth textures, perfect lighting and unnatural facial details. The next-generation system is said to address these limitations by producing images that more closely resemble real-world photography.
The new gpt-image-2 model is also expected to offer improved composition capabilities. The model is said to generate more structured visuals including accurate placement of text, objects and layouts. This may make it more useful for practical applications such as presentations, design mock-ups and visual storytelling.
Also read: Meta to train new workforce for data centres amid the layoff reports
In the meantime, ChatGPT is reportedly testing a new reference photo feature which may allow users to upload an image once and reuse it across multiple prompts. The feature is designed to streamline workflows by reducing the need for repeated uploads.
If true, the new model will directly rival Google and Anthropic’s advanced image generation capabilities. However, the details including release timeline, benefits and validity about the model still remain under the wraps.
Meanwhile, OpenAI is also working on a new GPT 5.5 model that is codenamed Spud. It is expected to launch next week and may bring big enhancements compared to existing models. Again, we may have to wait for a few more days to get official confirmation.
