• LemmysMum@lemmy.worldOP
      link
      fedilink
      English
      arrow-up
      7
      arrow-down
      1
      ·
      10 months ago

      Faces have a lot more acceptable variation, and far fewer configurations, that still results in something cohesive. Hands have far fewer acceptable variations and far more extensive positions and configurations that makes consistency difficult to model without specific training data.

        • LemmysMum@lemmy.worldOP
          link
          fedilink
          English
          arrow-up
          5
          arrow-down
          1
          ·
          edit-2
          10 months ago

          Some open, some closed. More variations of flowers from bud to petals than even faces, less positional requirements. So right place, lacking definition.

    • nudermeisters@lemmynsfw.com
      link
      fedilink
      English
      arrow-up
      5
      ·
      10 months ago

      We will get there quickly. We need a better intermediate AI that translates natural language to prompts, and also guides the image generation.