A. GUI Installation:
1. Doonload an’ run the dot executable for yer OS:
1.1. Windows (testit on Windows 10 an’ 11):
– Doonload dot.zip, unzip it an’ then run dot.exe
1.2. Mac (Testit on Apple M2 Sonoma 14.0):
– Doonload dot-m2.zip an’ unzip it.
– Open terminal an’ run.
Code: Select all
xattr -cr dot-executable.app
tae remove ony extended attributes.
In case o’ camera readin’ error:
– Right click an’ choose Show Package Contents
– Execute dot-executable frae Contents/MacOS folder.
A. How tae Use the GUI:
Example o’ how tae dae it:
1. Stick the original picture intae the bit marked ‘source’.
2. Write doon the camera ID nummer in the ‘target’ box. In maist cases, 0 is the right yin.
3. Write doon the config file in the ‘config_file’ bit. Pick a default setup frae the dropdown list or use yer ain file.
4. (Optional) Tick the ‘use_gpu’ box if ye want tae use the GPU.
5. Click the ‘RUN’ button tae get the deepfake goin’.
B. CLI Instawlation:
1. Install Pre-requisites:
Linux:
Code: Select all
sudo apt install ffmpeg cmake
MacOS:
Code: Select all
brew install ffmpeg cmake
Windows:
a) Dounload an’ install Visual Studio Community.
b) Install Desktop development wi’ C++ frae the Visual studio installer.
c) Create Conda Environment.
d) The instructions assumes that ye hae Miniconda installed on yer machine. If ye dinnae, ye need tae install it.
2. Wi’ GPU Support:
Code: Select all
conda env create -f envs/environment-gpu.yaml
Code: Select all
conda activate dot
Install the torch an torchvision dependencies based on the CUDA version installed on yer machine:
Install CUDA 11.8.
Install cudatoolkit frae conda:
Code: Select all
conda install cudatoolkit=<cuda_version_no> (replace <cuda_version_no> with the version on your machine)
Install torch an torchvision dependencies:
Code: Select all
pip
install torch==2.0.1+<cuda_tag>
torchvision==0.15.2+<cuda_tag> torchaudio==2.0.2 --index-url
https://download.pytorch.org/whl/cu118, where <cuda_tag> is the
CUDA tag defined by Pytorch. For example, pip install torch==2.0.1+cu118
torchvision==0.15.2+cu118 torchaudio==2.0.2 --index-url
https://download.pytorch.org/whl/cu118 for CUDA 11.8.
* Note: torch1.9.0+cu111 can also be used.
Tae check that torch an torchvision are installed correctly, run the following command:
Code: Select all
python -c "import torch; print(torch.cuda.is_available())".
If the output is True, the dependencies are installed wi’ CUDA support.
3. Wi’ MPS Support (Apple Silicon):
Code: Select all
conda env create -f envs/environment-apple-m2.yaml
Code: Select all
conda activate dot
Tae check that torch an’ torchvision are instawed richt, rin the followin’ command:
Code: Select all
python -c "import torch; print(torch.backends.mps.is_available())".
If the ootput is True, the dependencies are instawed wi’ Metal programmin’ framework support.
4. Wi’ CPU Support (slow, no’ recommended):
Code: Select all
conda env create -f envs/environment-cpu.yaml
Code: Select all
conda activate dot
Instaw the dot:
Code: Select all
pip install -e .
– Dounload Models.
– Dounload dot model checkpoints.
– Unzip the dounloaded file in the root o’ this project.
CLI Usage:
Run
Code: Select all
dot --help
tae get a full leet o’ available options.
C. SimSwap
Code: Select all
dot -c ./configs/simswap.yaml --target 0 --source "./data" --use_gpu
SimSwapHQ
Code: Select all
dot -c ./configs/simswaphq.yaml --target 0 --source "./data" --use_gpu
FOMM
Code: Select all
dot -c ./configs/fomm.yaml --target 0 --source "./data" --use_gpu
FaceSwap CV2
Code: Select all
dot -c ./configs/faceswap_cv2.yaml --target 0 --source "./data" --use_gpu
Note: Tae enable face superresolution, use the flag
Code: Select all
--gpen_type gpen_256 or --gpen_type gpen_512.
Tae use dot on CPU (not recommended), dinnae pass the “–use_gpu flag”.
1. Controllin dot wi the CLI:
Disclaimer: We use the SimSwap technique for the followin demonstration.
Runnin’ dot via ony o’ the abuin methods generat’s real-time Deepfake on the input video feed usin’ source images frae the data/folder.
1.1. Docker:
– Settin’ up docker.
– Build the container.
Code: Select all
docker-compose up --build -d
– Access the container.
Code: Select all
docker-compose exec dot "/bin/bash"
– Connect docker tae the webcam.
1.2. Ubuntu:
– Build the container.
Code: Select all
docker build -t dot -f Dockerfile .
Run the container.
Code: Select all
xhost +
docker run -ti --gpus all \
-e NVIDIA_DRIVER_CAPABILITIES=compute,utility \
-e NVIDIA_VISIBLE_DEVICES=all \
-e PYTHONUNBUFFERED=1 \
-e DISPLAY \
-v .:/dot \
-v /tmp/.X11-unix:/tmp/.X11-unix:rw \
--runtime nvidia \
--entrypoint /bin/bash \
-p 8080:8080 \
--device=/dev/video0:/dev/video0 \
dot
1.3. Windows:
Follow thae instructions unner Windows tae set up the webcam wi docker.
– Build the container.
Code: Select all
docker build -t dot -f Dockerfile .
– Run the container.
Code: Select all
docker run -ti --gpus all \
-e NVIDIA_DRIVER_CAPABILITIES=compute,utility \
-e NVIDIA_VISIBLE_DEVICES=all \
-e PYTHONUNBUFFERED=1 \
-e DISPLAY=192.168.99.1:0 \
-v .:/dot \
--runtime nvidia \
--entrypoint /bin/bash \
-p 8080:8080 \
--device=/dev/video0:/dev/video0 \
-v /tmp/.X11-unix:/tmp/.X11-unix \
dot
1.4. macOS:
Follow thae instructions here tae set up the webcam wi docker.
– Build the container.
Code: Select all
docker build -t dot -f Dockerfile .
– Run the container.
Code: Select all
docker run -ti --gpus all \
-e NVIDIA_DRIVER_CAPABILITIES=compute,utility \
-e NVIDIA_VISIBLE_DEVICES=all \
-e PYTHONUNBUFFERED=1 \
-e DISPLAY=$IP:0 \
-v .:/dot \
-v /tmp/.X11-unix:/tmp/.X11-unix \
--runtime nvidia \
--entrypoint /bin/bash \
-p 8080:8080 \
--device=/dev/video0:/dev/video0 \
dot
D. Virtual Camera Injection:
Instructions vary dependin’ on yer operatin’ system.
1. Windows:
– Install OBS Studio.
– Install VirtualCam plugin.
– Choose Install an’ register only 1 virtual camera.
– Run OBS Studio.
– In the Sources section, press on Add button (“+” sign).
– Select Windows Capture an’ press OK. In the appeared window, choose “[python.exe]: fomm” in Window drop-down menu an’ press OK. Then select Edit -> Transform -> Fit tae screen.
– In OBS Studio, go tae Tools -> VirtualCam. Check AutoStart.
– Set Buffered Frames tae 0 an’ press Start.
– Now OBS-Camera camera should be available in Zoom (or other videoconferencing software).
2. Ubuntu:
Code: Select all
sudo apt update
sudo apt install v4l-utils v4l2loopback-dkms v4l2loopback-utils
sudo modprobe v4l2loopback devices=1 card_label="OBS Cam" exclusive_caps=1
v4l2-ctl --list-devices
sudo add-apt-repository ppa:obsproject/obs-studio
sudo apt install obs-studio
– If ye cannae find tools –> v4l2sink in OBS Studio, follow these steps:
Code: Select all
mkdir -p ~/.config/obs-studio/plugins/v4l2sink/bin/64bit/
ln -s /usr/lib/obs-plugins/v4l2sink.so ~/.config/obs-studio/plugins/v4l2sink/bin/64bit/
Usin’ the virtual cam wi’ OBS Studio:
– Open up OBS Studio.
– Go tae tools –> v4l2sink
– Pick /dev/video2 an’ YUV420
– Click “start”.
– Join a meetin’ an’ choose “OBS Cam”.
3. macOS:
– Dounload an’ install OBS Studio fir macOS.
– Open OBS an’ follow the first-time setup (ye micht need tae enable certain permissions in System Preferences).
– Run dot wi’
Code: Select all
--use_cam flag
tae enable camera feed.
– Click the “+” button in the sources section → select “Windows Capture”, create a new source an’ enter “OK” → select windae wi’ “python” included in the name an’ enter OK.
– Click “Start Virtual Camera” button in the controls section.
– Select “OBS Cam” as default camera in the video settings o’ the application target o’ the injection.
4. Run dot wi’ an Android emulator:
If yer daein’ a test agin a mobile app, virtual cameras are a wee bit trickier tae inject. An alternative is tae use mobile emulators an’ still resort tae virtual camera injection.
– Run dot. Check runnin’ doot fur mair info.
– Run OBS Studio an’ set up the virtual camera. Check virtual-camera-injection fur mair info.
– Dounload an’ Install Genymotion.
– Open Genymotion an’ set up the Android emulator.
4.1. Set up dot wi’ the Android emulator:
– Open the Android emulator.
– Click on camera an’ select OBS-Camera as front an’ back cameras. A preview o’ the doot window should appear. In case there’s nae preview, restart OBS an’ the emulator an’ try agin. If that didnae work, use a different virtual camera software like e2eSoft VCam or ManyCam.
– dot deepfake output should noo be the emulator’s phone camera.