In this notebook, I deepfaked myself by adapting from a demonstration from the paper First Order Motion Model for Image Animation. You can find more information here in their website.

A simple-to-understand video on the First Order Motion Model and what sets it apart from other contemporary deepfake techniques is discussed in this YouTube video [Everybody Can Make Deepfakes Now!](https://www.youtube.com/watch?v=mUfJOQKdtAk).

I was inspired from this YouTube video titled How It's Memed: First Order Motion Model Deepfake.

Google Colab demo, from which I adapted and modified from, can be found in their GitHub repo here. The pre-trained checkpoint (which is required but since it's too big thus I excluded it in this repo) can be obtained from this shared Drive here or, you can look at the repo's README.md to find it. Place the .tar file in the first-order-motion-model directory.

Note that the model checkpoint is trained on 256*256 pixel images. From my understanding, at time of this writing there's no other pre-trained sizes available unfortunately.