Following on from the FakeApp v1.1 tutorial (which should be used as a primer) – comes the newest version 2.1 – make sure you follow the instructions carefully, if you need help you can find it within the Reddit communities.
One-button video creation.
One-button dataset creation.
More streamlined, cohesive UI.
Packaged into one installer, potential for desktop/start menu shortcuts.
Abstracted out command prompts.
Packaged FFMPEG, removing the need for manual video-to-image conversion.
Text fields replaced with more intuitive drop-downs.
Split some videos with your two desired faces into two sets of a few hundred frames each with a tool like FFMPEG. If you use FFMPEG, the command you want is: ffmpeg -i scene.mp4 -vf fps=[FPS OF VIDEO] "out%d.png". After splitting, run both directories of split frames through the “Extract” tool to produce training data
Switch to the “Train” tool, and input the paths of the training data produced in step 1 (it should be in a folder called “aligned”) as well as the “models” folder along with this project (which you can move somewhere convenient)
Train until the preview window shows results you are satisfied with
Split the video to be faked into frames and run the “Convert” tool on them to create faked frames, which can then be re-merged into a deepfaked video
Copy and reuse the same encoders for faster results in future fakes
-CUDA 8.0 must be installed, and its bin folder must be included in the PATH environment variable.
-At least a few GB of free space on disk to allow the app to create Temp files
-Run fakeapp.bat to launch the app
–RuntimeError: module compiled against api version 0xc but this version numpy is 0xb is just a warning related to how the alignment libraries were installed, the app will run properly despite it appearing if no other errors occur
-It may take 30-45 seconds after pressing the Start button for the app to unpack and start the training/merging scripts the first time
-You can still quit training by focusing the training window and pressing “q”
-Paths to models/data must be absolute, not relative
Deepfakes FakeApp v1.1 Tutorial | TensorFlow Neural Network Face Replacement
This tutorial on how to create deepfakes on Windows is solid but his results are hilariously bad. This is only due to the amount of time Brambo dedicated to the machine learning (should have been a good few hours more than allowed in this clip).
This goes over using FakeApp v1.1. You can use the models generated through this app in FakeApp v2.0. You must have a windows machine, Nvidia GPU, and the ability to install CUDA 8.0. All other OS/GPUs are not supported.
Download CUDA 8.0 and store its bin folder in the PATH environment variable (when I installed Cuda this was automatically done for me, but if you have Cuda related errors then you should double check the path)
Google’s first result of ‘CUDA 8.0‘ will give you a link to the latest version of CUDA Toolkit (for Nvidia users only!).
There’s a new trend on the interwebs called ‘Deepfakes’, a machine learning system that can be trained to paste one person’s face onto another person’s body, complete with facial expressions.
The effect isn’t yet more convincing than conventional computer graphics techniques, but it could democratize Hollywood-level special effects fakery — and, potentially, lead to a flood of convincing hoaxes.
I’ll explain how DeepFakes works both programmatically and theoretically in this video. It’s essentially 2 autoencoders trained on 2 image datasets and then we reconstruct image A using image B’s decoder.
“I was going to tell a science fiction story about face-swapping, and mass blackmail. Then the news broke about unethical face-swapping videos, and software designed and marketed for creating them: and I realised the future had arrived faster than I thought.”
Over at Reddit, user deepfakes has released a program that allows users to create “deep fakes,” or nearly seamless manufactured images. As with about 99 per cent of all tech-related innovation, the first use and proof-of-concept have to do with porn. Specifically, the app allows you to face swap your favourite person’s face onto a porn actor’s:
In December, Motherboard discovered a Redditor named ‘deepfakes’ quietly enjoying his hobby: Face-swapping celebrity faces onto porn performers’ bodies. He made several convincing porn videos of celebrities—including Gal Gadot, Maisie Williams, and Taylor Swift—using a machine learning algorithm, his home computer, publicly available videos, and some spare time.
Back in December, the unsavoury hobby of a Reddit user by the name of deepfakes became a new centrepiece of artificial intelligence debate, specifically around the newfound ability to face-swap celebrities and porn stars. Using software, deepfakes was able to take the face of famous actresses and swap them with those of porn actresses, letting him live out a fantasy of watching famous people have sex.
Now, just two months later, easy-to-use applications have sprouted up with the ability to perform this real-time editing with even more ease, according to Motherboard, which also first reported about deepfakes late last year.
The faces of celebrities, politicians, children, or pretty much anyone, can be pasted over faces of porn stars in X-rated movies using freely available machine-learning software.
The resulting flicks look convincing, and effectively allow miscreants to place people – from the rich and famous to the powerful to ex-partners – into highly compromising and believable positions on moving video footage.
As if female celebrities don’t have enough to worry about in regards to creepy strangers on the internet, now people are making fake porn using their faces.
Motherboard reports that people keep making “deepfakes,” which are AI-assisted porn videos using celebrity faces. After a Redditor named deepfakes started releasing videos involving Taylor Swift, Maisie Williams, and Gal Gadot late last year using machine-learning, another Redditor was inspired to create an app to help other make similar fake porn.