Deepfakes Examples & Tutorials: AI in Media Production #AI

The idea of Deepfakes is scary. In layman’s terms, it’s using artificial intelligence to create a convincing face swap in video and imagery. Deepfakes can be used to mislead viewers into believing someone said something they didn’t, creating fake news or hoaxes, and worse. But, there is a big future for deepfakes in the film industry.

Photo credit: Matheus Ferrero @matheusferrero

Deepfakes: Should we still trust what we see?

This video from The Telegraph explains what Deepfakes are in a very easy to understand way.

While this technology can also be used for very bad things, it can also be used for good, or for entertainment, at least. The positive use of this technology is what we’re focusing on in this AI series.

The uses of this technology in the film industry are far-reaching. AI can make actors completely immortal. Think of the Star Wars films where Grand Moff Tarkin was digitally recreated after actor Peter Cushing died. AI could make that task much simpler. De-aged actors in flashbacks or time travel scenes, for example, can look more realistic with these tools. Virtual characters will be able to do things that real humans can’t do in and go places they can’t go, without greenscreen. Coupled with traditional VFX, we are in for some astonishing visuals on the big screen. Likewise, consider the use of this type of AI in video games and VR. Certainly, it’s endless!

Without question, major ethics questions will emerge here, but we are solely focusing on technology. If technology can get past the uncanny valley, the sky is the limit.

How will Deep Fakes Affect Filmmaking?

• How Artificial Intelligence Will Make Digital Humans Hollywood’s New Stars
• Deepfake Tech Eyed by Hollywood VFX Studios
• Deepfake And What It Could Mean For The Film Industry
• Deepfakes could destroy trust in filmmakers. What can we do about it? – RedShark News, 4 October 2019

How Deepfakes Work: Generative Adversarial Networks

Generative Adversarial Networks (GANs) is a machine learning system that generates realistic images from sketches, images, or videos. The computer learns the images. Rob Miles explains GANs in a non-computer programmer way.

A Style-Based Generator Architecture for Generative Adversarial Networks

Learn more about GANS

• Google Machine Learning
• Generative Adversarial Networks: The Tech Behind DeepFake and FaceApp
• Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow
• Deep learning: Amazing feats with Generative Adversarial Networks (GANS)

This AI Learned How To Generate Human Appearance

This series explains how NVIDIA’s AI created high-resolution images of faces and human bodies using GANs.

Technology Vs. Creativity: Digital Humans, THU 2019

The CG Garage is a podcast from The Chaos Group Labs. This episode is from the panel at “Trojan Horse was a Unicorn”, or THU 2019, in short. It focuses on digital humans. Christopher Nichols talks to Digital Rembrandt Ian Spriggs, CG pioneer Paul Debevec, ILM’s chief re-animator Hal Hickel, acclaimed anatomy artist Scott Eaton, and Naughty Dog lead character artist Frank Tzeng.

Examples of Deepfakes

No, this is not all useful but it shows you the power of the technology!

I used “deepfakes” to fix the Lion King

Nikolay Mochkin used Deepfakes to make the new Lion King look more like the original animated film, and it’s a lot of fun.

The Shining starring Jim Carrey: Episode 3 – Here’s Jimmy! [DeepFake]

Jack Nicholson’s face is flawlessly replaced with Jim Carrey’s face in The Shining. There are loads of these types of videos on YouTube, with Tom Holland playing Marty McFly, Keanu Reeves as Forrest Gump, and more.

Deepfakes Tutorials

This tutorial series uses NVIDIA GPU. Get Deepface Lab at Github.

Lesson 2 goes deeper into the topic.