Summary

  • AI image generators have improved drastically over the past couple of years, but they still often display tell-tale signs of manipulation.
  • Some of these include eyes that look eerily vacant or expressionless, mangled or mismatched branding, and reflections that contain distortions or extraneous elements not present in the original image.
  • Another common flaw is the absence of depth, with reflections and shadows often lacking.
  • Additionally, backgrounds are often confusing, with objects or architecture making little sense when taken in context with the rest of the image.
  • Text can often be garbled, with strokes that resemble letters but don’t form actual words.
  • While these flaws are less prevalent in newer models, the underlying issues still tend to show up in one way or another.

By Tyler Fyock

Original Article