• Emerald@lemmy.world
    link
    fedilink
    English
    arrow-up
    5
    arrow-down
    2
    ·
    edit-2
    9 months ago

    Consistency is still an issue. It’s hard to generate multiple images or videos and have a consistent visual style with ai

    • UraniumBlazer@lemm.ee
      link
      fedilink
      English
      arrow-up
      3
      arrow-down
      2
      ·
      9 months ago

      Not necessarily. Fine-tuning models can solve this issue to a great degree. The model’s behavior is largely dependent on its training data. If it has generic training data, it’s going to produce generic images.

      See Corridor crew’s anime experiment. They managed to solve this issue to a great degree in their second version. It’s quite cool!