This course shows the current state of the art of future-oriented technology machine learning with a specific application, namely the exchange of faces in a video. This process is also known as deepfake.
Deepfake videos are almost indistinguishable from real ones. In contrast to photos, until recently it was very time-consuming to manipulate videos in a photo-realistic manner. But that has been over at the latest since the triumph of machine learning:
AI-based tools can also be used almost unnoticed in the moving image and exchange faces or put them on other bodies.
We use the current tool DeepFaceLab step by step to change the face in a video, so that is looks deceptively real.
And we also take a look at Avarify (based on the First Order Model): Control Photorealistic avatars for video-conferencing in realtime.
Are there any requirements or requirements for the course?
No programming skills are required
For one part of the course a mid-range Nvidia graphics card is recommended ( min. 4 GB VRAM)
(only for DeepfaceLab, for Avatarify you can render the code remotely on a google server :-)
Who is this course for?
Beginners interested in machine learning - deepfakes
Advanced developers who want to generate their own deepfakes
A view of what awaits you:
Step-by-step with DeepFaceLab, we go through everything together
Download the software
Preparation of the data
Training the model
Exchange of faces
Video generation and review
Avatarify - Realtime Face-Swap, where your control any Avatar (only 1 Picture needed) with your Webcam
The course contains a concrete practical examples.
You not only learn how to create deepfakes,
but also how you apply it directly.
I´m not advocating the usage of Deepfake technology for illegal or unethical activities.
Please only use in the legal frame. There is one lession about this point in the course.