OpenAI Sharpens ChatGPT’s Image Tools With Faster Output, Cleaner Edits

OpenAI on Tuesday rolled out a new version of ChatGPT Images, powered by its latest flagship image generation model, promising faster performance and more precise edits while preserving key visual details.

The updated model can generate images up to four times faster and is designed to better follow user instructions, whether creating images from scratch or editing uploaded photos. 

According to the company, the system can now make targeted changes without altering elements such as lighting, composition, or a person’s appearance unless explicitly requested.

OpenAI is also introducing a dedicated Images space within ChatGPT to make image generation more intuitive and exploratory. The feature includes preset styles, filters, and visual ideas to help users experiment without writing detailed prompts.

The new Images model is rolling out globally today for all ChatGPT users and is also available via the API as GPT Image 1.5. Most users will see the update immediately, while access for Business and Enterprise customers will follow later.

Image editing is a central focus of the upgrade. The model can now more reliably add or remove objects, blend elements, transpose visuals, and apply stylistic changes while preserving the core of the original image. Use cases range from photo touch-ups and virtual clothing try-ons to more conceptual visual transformations.

The system also shows improvements in instruction following, text rendering, including smaller and denser text, and handling complex scenes such as images with many faces, producing more natural-looking results.

The upgrade narrows the gap with Google’s Imagen models, which excel at photorealistic generation inside Gemini but offer fewer fine-grained editing controls in consumer-facing tools.

For developers, GPT Image 1.5 brings the same improvements to the API, with stronger image preservation than the earlier GPT Image 1. OpenAI said this makes it better suited for brand assets, marketing visuals, and generating large product image catalogs from a single source image.

The company has also reduced image input and output costs by 20% compared to the previous version, allowing teams to iterate more within the same budget.

The company said image generation remains an evolving area and that while the model shows clear gains, results are not yet perfect. The earlier version of ChatGPT Images will remain available as a custom GPT.

Future updates will focus on finer-grained edits, richer visual detail, and improved performance across languages, OpenAI said.