TOOL · Image Generation
GPT-Image 2
A short field note on GPT-Image 2 seen through one narrow lens, how it renders sheer black stockings. Three generated portraits of the same woman turning away from the camera, posted by someone who says the result stopped him and that the material came through better than he expected.
Here's the thing.
@you1873118 (serein) posted three images and four short lines, and the whole thing hinges on a single moment. The moment she turns around.
His words, he looked over and froze. It got to him enough that he went back and asked the AI whether it was really her.
The tool behind it, chat-image2.
What he singles out is narrower than you'd expect for a post like this. Not the face, not the lighting, the sheer black stockings. In his words, the model renders them to the fullest.
That's the note worth keeping. It's a small, specific read on where image models currently land on fabric and material realism, which is exactly the kind of detail that used to give generated images away.
At the time of writing, the post sits at 87 likes, 31 bookmarks and 6.5K views. Original thread here, x.com/you1873118/status/2096460410711118080
