Neural network animating images unveiled
MagicAnimate neural network enables animation of any photo or image (Collage: RBC-Ukraine)
An open-access neural network called MagicAnimate has been released, capable of generating animations based on a provided image. It is gaining popularity due to its realistic interaction with humans: facial expressions and movements take on a frighteningly realistic character, reports GitHub.
The neural network can animate the Mona Lisa doing yoga. You can also try on popular dance moves from social networks without getting out of bed.
The process is simple: you need to upload a photo of the model and a video reference of the movement, after which you only have to wait a few minutes. The neural network animates several people at once and does not perform a check for copyright on images or videos.
Neural network automatically adds missing elements to objects (GIF: GitHub)
This project is a joint initiative of the National University of Singapore and ByteDance, the parent company of TikTok. Anyone can try out the neural network by using the demo version on the Hugging Face platform.
However, at the moment, the servers are experiencing high traffic, and it is necessary to periodically retry attempts, keeping in mind that MagicAnimate does not offer paid options at the moment.
