Google will release an artificial intelligence model called Lumiere that can turn text and images into realistic, editable videos.

The model uses space-time diffusion and space-time U-Net architecture, which allows it to instantly generate video in a single pass. Users can enter text descriptions or upload images to create dynamic videos.

While other AI-powered video generators already exist, the researchers say Lumiere's approach to time-sampling data is new. The model was trained on a large data set, but there was no mention of where the data came from. This raises questions about copyright infringement.

Previously, lawsuits have been filed against generative model developers for misuse of content in the learning process. #google #googleai #ai