OpenAI launches ChatGPT Images 2.5 with more accurate editing and faster generation
OpenAI has unveiled ChatGPT Images 2.5, offering better instruction following, precise single-element edits, more realistic output and 50% faster generation.
The new model is the latest update to image generation in ChatGPT. Android Authority said ChatGPT Images 2.5 maintains key details from reference images and can make targeted changes to a single element in a photo. It also supports editing over multiple steps without degrading quality, a capability intended for more complex corrections.
OpenAI said in a blog post that images generated by the model have more natural lighting and textures, and that the system avoids wiping out subjects when an uploaded image is edited. OpenAI also claims ChatGPT Images 2.5 is 50% faster at producing images than previous versions.
The release is part of OpenAI's continuing push to make image creation in ChatGPT easier and more realistic. According to Android Authority, users have long used ChatGPT for image generation and editing, sometimes seeing odd outputs; the new model is designed to make the process more controllable and less hit-or-miss.