OpenAI says ChatGPT Images 2.5 brings sharper details, more natural lighting, richer textures, and better preservation of people and objects from reference photos. The company also says it can generate images up to 50% faster and follow complex visual instructions more reliably.
The model adds Sketch for turning rough drawings into finished images, templates for posters and product photography, and comments directly on images during edit requests. Early Reddit reactions, however, have focused on visible noise, with users describing the patterns as awful and some results as terrible.
I tested four images: a mug in a studio, a barn owl at twilight, a fantasy scene with a winged horse and dragon, and two friends on a rooftop. The model handled detailed subjects well, including feathers, facial expressions, and clothing. But fine grain appeared repeatedly in skies, studio backgrounds, hair, and skin, even when prompts requested smooth gradients and no visual noise.
The texture does not ruin every image, but it can make results look overprocessed and weaken the realism that OpenAI promises. ChatGPT Images 2.5 is fast and capable, yet its persistent granular texture remains a noticeable limitation in these tests.
Comments
0No comments yet. Be the first to comment.