you are viewing a single comment's thread
view the rest of the comments
[+] [S] -1 points 2 months ago (4 children)
  • [–] -3 points 2 months ago (3 children)

    If I had the right GPU, I'd be tempted to go back and use 2023-ass video generators to adapt an original story, just to demonstrate - no tool is immune to human expression. Of fucking course you can convey complex feelings despite visible constraints. Anyone going 'well why didn't you just animate it?' is incorrect, in the same way as if they'd told an animator, 'well why didn't you just film it?'

  • source
  • parent
  • hideshow 3 child comments
  • [+] [S] -1 points 2 months ago (2 children)
  • [–] -3 points 2 months ago (1 child)

    Corridor Digital seems to get it. They've dorked with it to fill in for missing frames. They're using it to relight flat greenscreen shots for their D&D... podcast? Anyway those relit shots look like PS3 cinematics, but they're putting real dudes in real costumes inside a video game, basically in real time. Nitpicking the quality is missing the point. Getting all the shots they're doing, but "properly," would take ten times as long and cost a hundred times as much. It's the same dudes putting in just as much work as the previous season. Anyone shouting 'not art!' doesn't mean things when they say words.

    The same dudes also used neural networks to fix greenscreen. They rendered a bunch of scenes with complex alpha blending, and built a neural network on that synthetic data. Unsurprisingly, the example-based universal approximator is pretty good at approximating from similar examples.

  • source
  • parent
  • hideshow 1 child comment
  • load more comments (1 reply)