v2 Tutorial - Intel-based Python (credits to MI7MHARX).

THIS IS FOR INTEL-BASED PYTHON, BUT YOU STILL CAN TRY TO ADAPT IT USING REGULAR PYTHON OR ANACONDA

RECOMMENDED FOR WINDOWS 10, WITH NVIDIA

this tutorial is made to help the beginners and the community. if you are quietly a pro, just do it by yourself.

link will be available within 2 hour

before we start,

install

    Intel-Python (w_python3_pu_2018.1.021) to C:\IntelPython3
    CUDA (cuda_8.0.61_win10)
    Visual C++ Build Tools (visualcppbuildtools_full)
    CMake (cmake-3.10.1-win64-x64)
    GIF Animator (GIFAnimator-Setup)

extract

    CUDNN to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0

setting the environment

    create and set PYTHONPATH to C:\IntelPython3 in Environment Variables

gear up!

in cmd, type:

conda install pip
pip install tensorflow-gpu dlib opencv-python keras scipy numpy h5py matplotlib tqdm scikit-image
conda install -c peterjc123 pytorch cuda80

(if the pip doenst work, download the wheel using the link in notes given, pick based on your system and python, then run pip install whatever_the-name-is.whl in the directory of the wheel downloaded)

intro

    get ready
    open code.txt in \face-swap
    run cmd, cd [FACE-SWAP DIRECTORY HERE]
    run in cmd, activate tensorflow and python train.py
    wait for several hours, stop the training by clicking on the faces window (not X), and then press Q.
    exit the cmd

get the data!

    collect tons of images of target (200+)

    gather the images in a new folder, and rename as target, and copy to \face-alignment-master

    find the video that satisfies your imagination (POV view recommended)

    convert the video to jpg by using FFMPEG or any video software

    to convert, read the code in bin directory in FFMPEG folder

    run cmd,

cd [BIN DIRECTORY INSIDE OF FFMPEG HERE] ffmpeg -i file.mp4 -r 1/1 $filename%d.jpg (change file.mp4 to your video name with its extensions, and $filename%d.jpg to any name eg riley%d.jpg)

    copy the jpgs to a new folder, rename to source, and copy to \face-alignment-master

i'm ready!

before do anything:

pip install -r requirements.txt
python setup.py install

    align both target and source images, to get aligned, cropped faces:

python align_images.py target python align_images.py source

    open target and source folder, take a look at the aligned folder in both folder
    remove unwanted images. rename it to targetA and sourceA. copy aligned folder in both target and source to \face-swap\data

    rename cage and trump to cageA and trumpA, and rename targetA and sourceA to cage and trump

    after done, copy align_images.py, merge_faces.py, umeyama.py, source folder to \face-swap

lets train!

    run the same code for training, activate tensorflow and python train.py
    after it done, new output folder can be access
    have a look of the training data. if satisfied, proceed
    run cmd python merge_faces_masked.py source
    a new aligned folder can be access in source folder
    take a look at the merged faces

but, no gif?

from the merged folder,

    go to GIFMaker , to convert jpg to gif. (it easier!)
    you can also use the install GIF Animator to make GIF
    DONE

kinda bored. need some audio!

use FFMPEG, to convert gif to video

ffmpeg -i try.gif -movflags faststart -pix_fmt yuv420p -vf "scale=trunc(iw/2)*2:trunc(ih/2)*2" try.mp4

use any video software

*add the source video (video + audio) *add the final deepfake video *align it correctly to the time/frames of the source video *delete the source video *render and voilà, now you have a new fake video with sound!

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

many thanks to deepfakes and community for this project