• Chozo@fedia.io
    link
    fedilink
    arrow-up
    7
    arrow-down
    4
    ·
    2 days ago

    This is inaccurate, current models can target specific zones in an image. Midjourney was capable of that ages ago, and Midjourney is already basically ancient history at this point.

    • kionay@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      10 hours ago

      this is correct, it’s called in-painting. you mat-out the area you want replaced and with modern GPT architecture you can also use text input to steer what gets in-painted

      not saying that that was for-sure done here since the changes don’t look too localized, but the technology absolutely exists