KlingAIResearch/LivePortrait ? reverse-engineered prompt

I want to build an app that can animate still photos, based on the popular LivePortrait project. The user experience should be simple: you upload a source image, like a portrait of a person or even a cat or dog. Then you upload a separate "driving video" that provides the motion, like someone talking or moving their head.

The app should then process these two files and generate a new video where the still photo comes to life, mimicking the movements and expressions from the driving video. It's important to have a clear option to choose between animating humans and animating animals, since they require different models. It should have a simple web UI for uploading files and viewing the result. Please handle all the backend setup, Python dependencies, and downloading the correct pre-trained models.