Qwen Image 2.1 draws criticism over anatomy and noisy images
A user testing the image model reports repeated problems getting a gun held correctly and describes its style as overly detailed and noisy.
TLDR
A user testing Qwen Image 2.1 describes its anatomy issues as far worse than those in Z-image, Krea2 or Illustrious, and compares its detailed, noisy look to GPT Image 1.5. They report repeated trouble getting a gun held correctly. A cyberpunk prompt fared better, but they say many background details still didn't make sense. Their broader argument is that benchmark-focused models prioritize text rendering and prompt following at the expense of anatomy and aesthetics.
Combined views
—
1 Source, first seen ago
— likes— comments— saves— reposts
