Deepfake can be described as the modern day answer to photoshop that uses artificial intelligence (or deep learning) to make fake images. It is essentially a synthetic media where a person in an existing image or video is replaced with someone else matching the likeness.
In the past, we have seen deepfake being used to make Jon Snow apologise for the dismal ending to Game of Thrones, Mark Zuckerberg talk about his total control over billions of people’s stolen data, and Barack Obama call Donald Trump a “complete dipshit“. With deepfake, also called manipulated media, you can put new words in the mouth of a politician or become a superstar.
Deepfakes have been used in financial fraud, hoaxes, and to spread fake news. Governments around the world are trying to detect and limit their usage while companies like Adobe are building tools that make it easier to detect them. However, if you are interested in testing them out for fun or for research, here are a few popular deepfake apps and tools. Please don’t use the technology to violate anyone’s privacy.
FaceApp needs no introduction in the deepfake world since the app developed by a Russian company called Wireless Lab is responsible for all the spirited communication around the subject. The app uses neural networks to generate highly realistic transformations of faces in photos.
The app is capable of transforming a face to make it smile, frown, look younger, or even change gender. It offers impression filters and ability to adjust temperature saturation. The app is also free on both App Store and Google Play Store and the free version puts a watermark on the final image.
DeepFaceLab is one of the leading softwares being used for creating deepfakes right now. Available for free on Windows and hosted on Github, it has become extremely accessible and led to a number of tutorials on the service online. It again uses neural networks to replace faces in videos and is being used by popular YouTube channels.
One of the biggest limitations with DeepFaceLab is that the service needs a certain amount of technical knowledge. As soon as you download, there is a need to unzip the software and you will see numerous folders and batch files. It even has a folder called ‘workspace’ consisting of all the training models. It also works with specific file names and locations in order for the batch file to function properly.
FaceSwap is an open-source deepfake system similar to DeepFaceLab that is available on Mac, Windows, and Linux for free. In comparison to DeepFaceLab, it provides more features, better documentation, and improved online support. It does this by auto-saving your training model but does require a modern GPU with CUDA for best performance.
With an active community, FaceSwap is powered by Python, Keras, and Tensorflow, and the software is actively developed. It allows users to import an initial video to produce a final deepfake with very limited interaction. You will be able to find a number of tutorials related to the platform.
While there are a number of deepfake videos out there, the one that brought the urgency was a video of US House Speaker Nancy Pelosi. The video slowed down her speech to make it look modulated. Zao does allow the option to modulate the voices of celebrities and stitch your face onto an actor’s body.
The difference being that Zao’s final output is meant to be fun and your friends should be able to identify it being fake with little effort. It is meant to help users mimic the personality of their favourite celebrity but users cannot save or take screenshots. The app is among the most downloaded free apps in China and gives users a ton of options across videos, outfits, and options are endless.
If you ever found a meme or a GIF making the wave and wondered if you could put your own face onto that meme, then Reface is the app for you. The app can swap your face with that of a celebrity or any other popular meme within a few seconds. With this app, you can play with live face swaps or gender swaps.
One of the biggest problems with this app is that it often keeps crashing and sometimes even fails to save the result. In simple terms, the result produced by Reface is really surprising and its developers claim that the image is deleted from the servers right after it is processed.
While the likes of FaceApp, Zao, DeepFaceLab dominate the deepfake ecosystem as either being the most popular or most used service, they are not the only services out there. Deepfakes web is another fast and easy to use tool for creating deepfakes, but it costs $2 for an hour. There is also FacePlay, which is available as a free service but users need to pay to unlock premium features.
One of the tools that you must check out is Adobe’s Project Morpheus, which was introduced last year to demonstrate the advantages and disadvantages of integrating deepfake technology into its products. While the initial results are far from being flawless, Adobe is looking to make its video editing tool capable of modifying users’ appearance with smooth and consistent outcomes.