I expect this is going to be the future of animation, since drawing in-between frames and backgrounds is very labor-intensive and expensive.
It'll be interesting to watch how copyright law treats training data. Suppose I want to make an animation in the style of Disney or Studio Ghibli or Kyoto Animation, and so I use their entire body of work as training data to generate output in the same style from my own sketches. Is that now a derivative work? Is it different when a human copies a style while drawing by hand (requiring a degree of effort and skill that few possess) versus if a computer does most of the work (requiring low effort and ordinary artistic skills)? Would animation be treated any differently than, say, training an AI to write songs like Bob Dylan or write stories like J.R.R. Tolkien or host a radio show like Garrison Keillor?
It'll be interesting to watch how copyright law treats training data. Suppose I want to make an animation in the style of Disney or Studio Ghibli or Kyoto Animation, and so I use their entire body of work as training data to generate output in the same style from my own sketches. Is that now a derivative work? Is it different when a human copies a style while drawing by hand (requiring a degree of effort and skill that few possess) versus if a computer does most of the work (requiring low effort and ordinary artistic skills)? Would animation be treated any differently than, say, training an AI to write songs like Bob Dylan or write stories like J.R.R. Tolkien or host a radio show like Garrison Keillor?