Creating a Deep Fake in under 10 lines of code on the cloud.

“What you see is Fake, What you hear is False.” -Somebody

The time is now. Anybody with no knowledge of AI or ML can now create a deep fake with just one image to place and a video to be faked.

Recently I have come across a video on YouTube by Two Minute Papers about Deep Fakes and I am intrigued by the pace of development in this technology. Until now Deep Fakes are complex to develop, need GPU and lots of computation power, and are time-consuming. Thanks to First Order Motion Model. I have tried the code given by @AliaksandrSiarohin on Github.

There are few conventional approaches to creating a deep fake, with Deep Learning(CNNs), GANs (Generative Adversarial Networks), Face Swap(specifically for faces), and now First Order Motion Model.

Sample Outputs :

Google Colab Demo Link :

Link: Google Colab, You can follow the instructions given in the notebook to run on your data.

Installation and Running on the local machine :

• Clone the repository

`git clone https://github.com/AliaksandrSiarohin/first-order-model`

• Install Requirements.

`pip install -r requirements.txt`

• Add/download drive folder https://drive.google.com/drive/folders/1kZ1gCnpfU0BnpdU47pLM_TQ6RypDDqgw?usp=sharing. To use your data, just change the path to your mounted drive path

To run a demo, download checkpoint and run the following command:

`python demo.py --config config/dataset_name.YAML --driving_video path/to/driving --source_image path/to/source --checkpoint path/to/checkpoint --relative --adapt_scale`

Sources :

Original repository: https://github.com/AliaksandrSiarohin/first-order-model

Two Minute Papers Youtube Video: https://www.youtube.com/watch?v=mUfJOQKdtAk&t=17s