Neuralens
Computer vision enabled checkout with built in object detection. One snap is all it takes.
Tried various fine-tuned models on clips from The Office. Many animations already use live-action as reference material, this shortens the process dramatically. Storyboard to live-action also near instant.
Interesting to see how the models understand the user's input words and tries to apply styles in novel forms. Peppa the pig (bottom right) doesn't contain human characters but it still seemed to know to skip the snout.
Controlnet seems able to generate background aspects as well as the subject given an initial input image.
