Coming up at SIGGRAPH 2021…
StyleCariGAN… “a caricature generation framework based on StyleGAN, which automatically creates a realistic and detailed caricature from an input photo, with optional controls on shape exaggeration and color stylization.”
AgileGAN, apparently an instant cartoon-ifer, less twisted than StyleCariGAN.
Auto-dance timed to the music beats and speed etc. Doesn’t the MikuMiku type software do this already? Well, if it doesn’t, it will soon.
Autocomplete for inking artists, ‘making several short curves to form a longer one’.
Smoke-ring fractal generation. Not exactly Gandalf-quality, but perhaps they can add modifiers for more complexity.
Realistic melting wax animation simulator.
Loose-mesh fabrics that behave realistically and can be ‘pulled’ onto a figure-part, like knitted socks being pulled onto feet.
Generate “real-world” ecosystems with balanced ecologies. Not sure if the animals get to eat each other.
Identify “internal structures” in a picture, auto-animate them as loops.
Better automatic pallette re-sorting.
Seamless manga inpainting, seemingly “for the first time”. Apparently to be used for auto-removal of speech-bubbles from old paper comics that are not in digital format. Not sure it’s that useful though, as mostly you’d just want to blank rather than remove the bubbles.
No advances in auto-coloring of greyscale paintings, by the look of it.